AU2012340186A1

AU2012340186A1 - Markers of triple-negative breast cancer and uses thereof

Info

Publication number: AU2012340186A1
Application number: AU2012340186A
Authority: AU
Inventors: Josh BAUER; Xi Chen; Brian LEHMANN; Jennifer A. Pietenpol
Original assignee: Vanderbilt University
Current assignee: Vanderbilt University
Priority date: 2011-11-18
Filing date: 2012-11-17
Publication date: 2014-06-19
Also published as: EP2780469A1; US20200224282A1; US20140303133A1; CA2856295A1; US20240158865A1; WO2013075059A1; US11788147B2

Abstract

In one aspect provided herein are methods of determining a triple negative breast cancer (TNBC) subtype in an individual in need thereof comprising determining expression of one or more genes in one or more TNBC cells of the individual; and comparing the expression of the one or more genes in the TNBC cells with the expression of the one or more genes in a control. In another aspect, the methods are directed to methods of determining a treatment protocol for the TNBC patient based on the TNBC subtype. In another aspect, the methods are directed to predicting whether an individual will benefit from a treatment for a particular TNBC subtype. In yet another aspect, the invention is directed to a method of determining whether an agent can be used to treat a TNBC subtype.

Description

WO 2013/075059 PCT/US2012/065724 MARKERS OF TRIPLE-NEGATIVE BREAST CANCER AND USES THEREOF RELATED APPLICATION This application claims the benefit of U.S. Provisional Application No.61/561,743, filed on November 18, 2011. 5 The entire teachings of the above application are incorporated herein by reference. GOVERNMENT SUPPORT This invention was made with government support under grants CA95131, CA148375; CA105436 and CA070856, CA68485 and CA009385 awarded by the 10 National Institute of Health (NIH). The government has certain rights in the invention. BACKGROUND OF THE INVENTION Treatment of patients with triple-negative breast cancer (TNBC), lacking estrogen receptor (ER) and progesterone receptor (PR) expression as well as human 15 epidermal growth factor receptor 2 (HER2) amplification, has been challenging due to the heterogeneity of the disease and the absence of well-defined molecular targets (Pegram MD, et al. J Clin Oncol. 1998;16(8):2659-2671; Wiggans RG, et al. Cancer Chemother Pharmacol. 1979;3(1):45-48; Carey LA, et al. Clin Cancer Res. 2007;13(8):2329-2334). TNBCs constitute 10%-20% of all breast cancers, more 20 frequently affect younger patients, and are more prevalent in African-American women (Morris GJ, et al. Cancer. 2007;110(4):876-884). TNBC tumors are generally larger in size, are of higher grade, have lymph node involvement at diagnosis, and are biologically more aggressive (Haffty BG, et al. J Clin Oncol. 2006;24(36):5652-5657). Despite having higher rates of clinical response to 25 presurgical (neoadjuvant) chemotherapy, TNBC patients have a higher rate of distant recurrence and a poorer prognosis than women with other breast cancer WO 2013/075059 PCT/US2012/065724 -2 subtypes (Haffty BG, et al. J Clin Oncol. 2006;24(36):5652-5657; Dent R, et al. Clin Cancer Res. 2007;13(15 pt 1):4429-4434). Less than 30% of women with metastatic TNBC survive 5 years, and almost all die of their disease despite adjuvant chemotherapy,which is the mainstay of treatment (Dent R, et al. Clin Cancer Res. 5 2007;13(15 pt 1):4429-4434). One of the first molecular insights into TNBCs was the observation that they are likely to arise in BRCA1 mutation carriers and have gene expression (GE) profiles similar to those of BRCA1-deficient tumors (Haffty BG, et al. J Clin Oncol. 2006;24(36):5652-5657). BRCA1 plays an important role in DNA double-strand 10 break repair, contributing to the maintenance of DNA stability (D'Andrea AD, Grompe M. Nat Rev Cancer. 2003;3(1):23-34). Poly ADP-ribose polymerase (PARP) enzymes are critical for appropriate processing and repair of DNA breaks (Hoeijmakers JH. Nature. 2001; 411(6835):366-374). Tumor cell lines lacking functional BRCA1 or BRCA2 are sensitive to PARP inhibitors in preclinical studies 15 (Farmer H, et al. Nature. 2005;434(7035):917-921). Clinical trials using both PARP inhibitors and DNA-damaging agents (e.g., cisplatin) in TNBC are currently underway and show promise in BRCA 1/2-mutant tumors (Fong PC, et al. N Engl J Med. 2009;361(2):123-134). Other studies identifying molecular markers of TNBC, such as VEGF (Burstein HJ, et al. J Clin Oncol. 2008;26(11):1810-1816), EGFR 20 (Nielsen TO, et al. Clin Cancer Res. 2004;10(16):5367-5374), Src (Finn RS, et al. Breast Cancer Res Treat. 2007;105(3):319-326), and mTOR (Ellard SL, et al. J Clin Oncol. 2009; 27(27):4536-4541) have been important for the design of clinical trials investigating targeted treatments. Clearly, there is a major need to better understand the molecular basis of 25 TNBC and to develop effective treatments for this aggressive type of breast cancer. SUMMARY OF THE INVENTION In one aspect provided herein are methods of determining a triple negative breast cancer (TNBC) subtype in an individual in need thereof comprising 30 determining expression of one or more genes (e.g., presence of one or more mRNAs WO 2013/075059 PCT/US2012/065724 -3 and/or protein encoded by gene) in one or more TNBC cells of the individual; and comparing the expression of the one or more genes in the TNBC cells with the expression of the one or more genes in a control. Increased expression of one or more genes comprising one or more cell cycle genes, cell division genes, 5 proliferation genes, DNA damage response genes or a combination thereof in the TNBC cells compared to the control indicates that the TNBC is a TNBC BLI subtype. Increased expression of one or more growth factor signaling genes, glycolysis genes, gluconeogenesis genes or a combination thereof in the TNBC cells compared to the control indicates that the TNBC is a TNBC BL2 subtype. Increased 10 expression of one or more immune cell signaling genes, cytokine signaling genes, antigen processing and presentation genes, core immune signal transduction genes or a combination thereof in the TNBC cells compared to the control indicates that the TNBC is a TNBC IM subtype. Increased expression of one or more cell motility genes, extracellular matrix (ECM) receptor interaction genes, cell differentiation 15 genes or a combination thereof in the TNBC cells compared to the control indicates that the TNBC is a TNBC M subtype. Increased expression of one or more cell motility genes, cellular differentiation genes, growth pathway genes, growth factor signaling genes, angiogenesis genes, immune signaling genes, stem cell genes, HOX genes, mesenchymal stem cell-specific marker genes or a combination thereof in the 20 TNBC cells compared to the control indicates that the TNBC is a TNBC MSL subtype. Increased expression of one or more hormone regulated genes in the TNBC cells compared to the control indicates that the TNBC is a TNBC LAR type. In another aspect, the methods are directed to methods of determining a treatment protocol for the triple negative breast cancer (TNBC) patient based on the 25 TNBC subtype. In another aspect, the methods are directed to predicting whether an individual (e.g., patient) will benefit from a (one or more) treatment for a particular TNBC subtype. In yet another aspect, the invention is directed to a method of determining 30 whether an agent can be used to treat a triple negative breast cancer (TNBC) subtype comprising contacting a cell line that is a model for the TNBC subtype of interest WO 2013/075059 PCT/US2012/065724 -4 with an agent to be assessed; and determining viability of the cell line in the presence of the agent, wherein if the viability of the cell line decreases in the presence of the agent, then the agent can be used to treat the TNBC subtype of interest. The TNBC subtype of interest can be a TNBC BL-1 subtype, a TNBC BL-2 5 subtype, a TNBC IM subtype, a TNBC M subtype, a TNBC MSL subtype or a TNBC LAR subtype. BRIEF DESCRIPTION OF THE DRAWINGS Figures IA-IC: Filtering GE data sets to identify TNBCs. (lA) Flow chart 10 of analysis. Human breast cancer GE profiles (training = 2353, validation = 894) were normalized within individual data sets and a bimodal filter was applied to select ER, PR, and HER2 negative samples by GE, resulting in 386 samples in the training set and 201 samples in the validation set with a triple-negative phenotype. k-means clustering was performed on the training set, and a GE signature 15 representing the TNBC subtypes from the training set was used to predict the best-fit subtype for each TNBC profile in an independent validation set. GSE-A was performed on the training and validation sets to identify enriched canonical pathways for each TNBC subtype. (IB) Histograms show the distribution and frequency of tumors using relative ER, PR, and HER2 GE levels (log2) and bimodal 20 fit to identify TN tumor samples. Dashed line indicates the expression value at the center of the positive expression peak used to select controls for C. (IC) Heat map representation of GE for 386 TNBCs relative to 5 IHC-validated controls for each ER, PR, and HER2. Figures 2A-2E: Identification of TNBC subtypes. (2A) Silhouette plot 25 showing the composition (n = number of tumors) and stability (AVG width) of k means clustering on the TNBC training set. Clusters with s(I)> 0 were considered stable. (2B) Consensus clustering displaying the robustness of sample classification using multiple iterations (x1000) of k-means clustering. (2C) The CDF depicting the cumulative distribution from consensus matrices at a given cluster number (k). (2D) 30 The optimal cluster number is 7 at the point in which the relative change in area (A) WO 2013/075059 PCT/US2012/065724 -5 under the CDF plot does not change with increasing k. (2E) Principal component analysis graphically depicting fundamental differences in GE between the TNBC clusters. The cluster (subtypes) are named as shown. Figure 3: GE patterns within TNBC subtypes are reproducible. Heat maps 5 showing the relative GE (log2, -3 to 3) of the top differentially expressed genes (P < 0.05) in each subtype in the training set (left) and the same differentially expressed genes used to predict the best-fit TNBC subtype of the validation set (right). Overlapping gene ontology (GO) terms for top canonical pathways in both the training and validation sets as determined by GSE-A are shown to the right of the 10 heat maps. Figures 4A-4F: Basal-like TNBC subtypes have differential sensitivity to DNA-damaging agents, IC50 values for TNBC cell lines treated with PARP inhibitors (4A) veliparib, (4C) olaparib, or (4E) cisplatin for 72 hours. Error bars reflect SEM for 3 independent experiments. Black horizontal lines above various 15 bars in the plots indicate cell lines that failed to achieve an IC50 at the highest dose of veliparib (30 ptM), olaparib (100 ptM), or cisplatin (30 pM). Cell lines that carry BRCA1 or BRCA2 mutations (pink) are displayed below the graph. Dot plot shows the log distribution of drug sensitivity to PARP inhibitors (B) veliparib, (4D) olaparib, or (4F) cisplatin in the basal-like subtypes (BL = BLI + BL2), the 20 mesenchymal-like subtypes (ML = M + MSL), and the LAR subtype, Black horizontal bars in the dot plot indicate the mean IC50 for each of the subtypes. *Statistically significant differences in IC50 values of BL compared with ML (P = 0.0 17) and LAR (P = 0.032), as determined by Mann-Whitney U test. Figures 5A-5D: Differential sensitivity of the LAR TNBC subtype to AR 25 and Hsp90 inhibitors. IC50 values for each TNBC cell line after treatment with (5A) bicalutamide or (5C) the Hsp90 inhibitor 17-DMAG for 72 hours. Black bar above bicalutamide indicates cell lines that failed to achieve an IC50. Heat map displays relative AR expression (log2) across TNBC cell lines. Dot plot shows log distribution of drug sensitivity to (5B) bicalutamide or (5D) 17-DMAG in the basal 30 like (BL = BL I + BL2), mesenchymal-like (ML = M + MSL), and LAR subtypes. Black horizontal bars in the dot plot indicate the mean IC50 for each of the subtypes.

WO 2013/075059 PCT/US2012/065724 -6 *Statistically significant differences in IC50 values of LAR versus BL (P = 0.007) or ML (P = 0.038) after bicalutamide and LAR versus BL and ML (P = 0.05) after 17 DMAG treatments, as determined by Mann-Whitney U test. Figures 6A-6D: Mesenchymal-like TNBC subtypes are sensitive to dasatinib 5 and NVP-BEZ235. IC50 values for each TNBC cell lined treated with (6A) dasatinib or (6C) NVP-BEZ235 for 72 hours. Cell lines that have PIK3CA mutations (red) or are deficient in PTEN (blue, circle indicates mutated) are displayed below the NVP-BEZ235 graph. Dot plots show the log distribution of drug sensitivity to (6B) dasatinib or (6D) NVP-BEZ235 in the basal-like subtypes 10 (BL = BLI + BL2), mesenchymal-like subtypes (ML = M + MSL), and LAR. Black horizontal bars in the dot plots indicate the mean IC50 for each of the subtypes. *Statistically significant differences in IC50 values of BL versus ML (P = 0.020) when treated with dasatinib and ML versus BL (P = 0.00 1) and LAR versus BL (P = 0.01) when treated with NVP-BEZ235, as determined by Mann-Whitney U test. 15 Figure 7: Xenograft tumors established from TNBC subtypes display differential sensitivity to cisplatin, bicalutamide, and NVP-BEZ235. Nude mice bearing established tumors (25-50 mm3) from basal-like (HCC 1806 and MDA-MB 468), LAR (SUM185PE and CAL-148), or mesenchymallike (CAL-51 and SUM159PT) were treated with cisplatin (red), bicalutamide (purple), NVP-BEZ235 20 (green), or vehicle (blue) for approximately 3 weeks. Serial tumor volumes (mm3) were measured at the indicated days. Each data point represents the mean tumor volume of 16 tumors; error bars represent SEM. Figure 8: Random distribution of TNBC subtypes found within dalasels. The percent of tumors in each cluster is displayed across 14 studies that comprise the 25 training dataset. Figure 9: Consensus clustering of training and validation TNBC datasets (n= 587). Heat map displays consensus clustering results depicting the robustness of sample classification. Red areas indicate samples that frequently cluster with each other over multiple iterations (1000) of k-means clustering (k::c 2 to k= 9). 30 Figure 10: Gene expression profiles derived from tumors that were laser capture microdissecled represent all TNBC subtypes. Bar graph depicts the WO 2013/075059 PCT/US2012/065724 -7 percentage of tumors that were microdissected (blue bars) present in the GSE584 (n=17) and Vanderbilt (n=14) datasets relative to non-microdissected tumors in the combined dataset (red bars). Figures 1 lA-I IB: TNBC subtypes identified by IHC. (1 1A) Silhouette plot 5 showing the composition (n= number of tumors) and stability (AVG width) of k means clustering on the TNBC training sel. Clusters with a silhouette width, s(i)>O were considered stable. (1IB) Heatmap displays hierarchical clustering of ERBB2, PGR, ESR1 and AR expression in the tumors identified by IHC. Color bar identifies the cluster associated with each tumor. 10 Figure 12: TNBC subtypes Identified by IHC display similar GE patterns to the six TNBC subtypes. IHC TNBC clusters are shown in silhouette plot with relative GE for genes unique to the six TNBC subtypes shown below. Figure 13: Differential GE across TNBC subtypes. Heatmaps show relative GE (log2, -2 to 2) associated with proliferation, DNA damage response, 15 myoepithelial genes, immune signal transduction, TGF3 signaling, growth factor receptors, EMT, WNT signaling, stem-like, claudin (CL), angiogenesis, AR and of AR target genes across TNBC subtypes (as in Figure 3). Figures 14A-14B: TNBC tumor subtypes differentially stain for the proliferation marker Ki-67. (14A) Representative micrographs from 20 tumors 20 showing IHC staining of the proliferation marker, Ki-67, in tumors from different TNBC subtypes. (14B) Dol plot showing the mean and distribution of Ki-67 staining within TNBC subtypes as scored by study pathologist. Subtypes : BLl= basal-like 1, BL2= basal-like 2, IM=immunomodulatory, M= mesenchymal-like, MSL= mesenchymal stem-like and LAR= luminal AR. 25 Figure 15: Response rates differ across TNBC subtypes in taxane-treated patients. Percent of patients achieving pathologic complete response (pCR) after taxane-bsed treatment was significantly (P=0.042, chi-squared analysis) different for patients whose tumors correlated to basal-like (n=19), mesenchymal-like (n=16) and luminal AR (n=7) subtypes, from studies in which response data were available. 30 Figure 16: IM subtype Is enriched in Immune cell signaling. Heatmaps showing the relative GE (log2, -2 to 2) for genes involved in immune cell surface WO 2013/075059 PCT/US2012/065724 antigens, cytokines, immune cell signal transduction, complement. chemokine, and antigen presentation across TNBC subtypes (as in Figure 3). Figure 17: The claudin-low predictor gene set identifies a sub-population of MSL tumors. Unsupervised hierarchical clustering was performed on the training 5 TNBC tumors using genes (n= 770) unique to the claud in- low subgroup [26]. Displayed to the right of the heatmap are the genes that are most differentially expressed (either high or low) in the claudin-low tumor set. Colorbar displays the TNBC subtype and heatmaps show relative levels (Log 2) 01 claudins (3, 4 and 7) and CD24, markers of this subgroup. 10 Figures 18A-18B: TNBC tumor subtypes differentially stain for AR by IHC. (1 8A) IHC staining for AR from 20 tumors with representative samples from each TNBC subtype shown. (1 8B) Dot plot showing the quantification of nuclear AR staining based on intensity and the percent of nuclei staining positive for AA in 20 tumors; Note, in some cases one dot represents overlapping dots from multiple 15 tumors. Figures 19A-19B:. An androgen-inducible gene signature segregates LAR tumors. Hierarchical clustering was performed on both the (19A) training and (19) validation TNBC tumor sel using a 559 androgen-inducible gene signature (Hayes MJ et al. Clin Cancer Res. 2008;14(13):4038-4044). 20 Figures 20A-20B: TNBC subtypes differentially correlate with the intrinsic molecular subtypes. (20A) Healmaps show relative GE (log2, -2 to 2) of luminal and basal markers of breast cancer across all TNBC subtypes. (20B) Bar graph shows the distribution of intrinsic molecular subtypes of breast cancer (luminal A and B, normal breast-like, HER2, basal-like or unclassified) within each TNBC subtype 25 (UNS= unstable, BLI=basal-like 1, B12=basal-like 2, IM= immunomodulatory, M=mesenchymal-like, MSL=mesenchymal stem-like and LAR=luminal AR) as determined by best-fit Spearman correlation to the intrinsic centroids. Figures 21A-21D TNBC subtypes differ in relapse-free survival and distant metastasis-free survival. Kaplan-Meier plot showing (21A) to-year RFS or (21C) 30 DMFS in TNBC subtypes. RFS (21B) and DMFS (21D) hazard ratios (bold) and WO 2013/075059 PCT/US2012/065724 -9 95% CI (parentheses) for patients from TNBC subtypes. Shaded boxes indicate significant (P < 0.05) comparisons. Figures 22A-22B: TNBC subtypes differ in age, but are similar size upon diagnosis. Box plots show the median (horizontal line), range (rectangle) and SD 5 (error bars) of (22A) age at diagnosis and (22B) tumor size (mm) between TNBC subtypes (as in Figure 3); - P = 9.0e-6 Figures 23A-23B: Chromosomal Aberrations in TNBC cell lines. (23A) Box plots depicting the average number of chromosomes (mode) in breast cancer cell lines correlating to basal-like (n=8) vs. mesenchymal-like (n=6) subtypes. (23B) 10 Box plots depicting the average number of chromosomal rearrangements (translocations, inversions, and deletions) in basal-like vs. mesenchymal-like subtypes. Chromosome mode and rearrangements were obtained from Departments of Pathology and Oncology, Univ of Cambridge (www.path.cam.ac.uk/-pawefish/cell%20line%20catalogues/breast-cell-lines.htm). 15 *P<0.01 by unpaired t-test. Figure 24: TNBC cell lines cluster in three major groups: luminal AR, basal like and mesenchymal-like. Unsupervised hierarchical clustering of TNBC cell lines performed on genes unique to TNBC subtypes (n= 2188). Colorbar shows the best correlation of cell lines to the TNBC subtypes. 20 Figures 25A-25C: LAR cell lines depend on AR expression for colony formation. (25A) Top panel depicts relative AR mRNA levels obtained from GE mlcroarrays performed on TNBC cell lines (log 2, centered on 0). Color bar identifies TNBC subtype classification for each cell line (immunomodulatory shown in orange and unclassified shown in red). Immunoblots showing relative expression 25 AR protein in TNBC cell lines, Hsp90 serves as a loading control. (25B) AR expression 72h following transfection with siRNA pools of non-targeting (NT) or targeting AR in MFM-223, MDA-MB-453 and SUM185PE cells. GAPDH expression serves as a loading control. (25C) Colony formation of MFM-223, MDA-MB-453 and SUM185PE cells 14d following siRNA transfection of AR or 30 non-targeting control (NT).

WO 2013/075059 PCT/US2012/065724 -10 Figures 26A-26B: Identification of PIK3CA mutations in AR-positive TNBC cell lines and tumors. (26A) Detection of PIK3CA 1047 mutation using digital droplet PCR. DNA from cell lines with varying PIK3CA status (HMEC = H1047R wt/wt, MDAMB453 = H1047R wt/mut, SUMI 85 = H1047R mut/mut and 5 SUM 159= H1047L wt/mut) were simultaneously amplified with fluorescent conjugated primers to wild-type (VIC) or mutant (FAM) and the percentage of wt vs. mut DNA was measured by digital droplet PCR. DNA that did not undergo PCR amplification is indicated by no reaction (NR). (26B) Heatmap displays the TNBC molecular subtype based of TCGA samples with corresponding levels of AR RNA 10 (RNA-seq) and protein (RPPA) for AR and phosphorylated AKT (T308 and S473). Color bar indicates PIK3CA mutations (red) within TNBC from the TCGA. Figure 27: AR-expressing TNBC cells display active P13K pathway. Heatmap displays relative protein levels (RPPA) of AR, AKT (S473 and T308), GSK3p (S9 and S21) and PTEN across a large panel of TNBC cell lines. 15 Hierarchical clustering was performed on cell lines and revel clusters of cell lines with active P13K through PTEN loss or PIK3CA mutations. Figures 28A-28C: AR-positive TNBC cell lines display active P13K pathway and are sensitive to P13K inhibitors. (28A) Immunoblot displays relative levels of AR, p-AKT, p-S6 in AR-dependent prostate cancer (LNCaP), human mammary 20 epithelial cells (HMEC) and LAR TNBC, with GAPDH serving as a loading control. (28B) IHC-stained sections display cellular levels of AR, and p-AKT in cell lines from the same tissue microarray (TMA). (28C) Bar graphs display the 50% inhibitory concentration (IC 50 ) for TNBC cell lines treated with single-agent pan P13K inhibitors (BKM-120 and GDC-0941) or PI3K/mTOR inhibitors (NVP 25 BEZ235 or GDC-0980) for 72 h. Black bars above graphs indicate cell lines in which the IC 50 was not reached at maximal dose. Bar below indicates normal HMEC cells and cell lines carrying PIK3CA mutations. Figures 29A-29B: Genetic targeting of AR increases the efficacy of the P13K inhibitor GDC-0941 and PI3K/mTOR inhibitor GDC-0980. (29A) Immunoblot 30 displays decreased AR expression at 72h following infection with two shRNAs targeting AR (shARI and shAR2) compared to infection with nontargeting shRNA WO 2013/075059 PCT/US2012/065724 - 11 (shNT) in three AR-positive TNBC cell lines. Actin levels below serve as loading controls. (29B) Line graphs display relative viability of LAR cell lines infected with nontargeting (shNT) or shRNAs targeting AR (shARI and shAR2) following a 72h treatment with the pan-PI3K inhibitor GDC-0941 (top) or the dual PI3K/mTOR 5 inhibitor GDC-0980. Figures 30A-30B: Pharmacological targeting of AR with bicalutamide increases the efficacy of GDC0941 and GDC0980 in AR-expressing TNBC. (30A) Line graphs show relative viability of AR-expressing cell lines treated with increasing doses of GDC-0941 (top) or GDC-0980 (bottom) alone or in combination 10 with 25pM bicalutamide (CDX). Dashed black line depicts the theoretical line of additivity of both drugs determined from the effect of bicalutamide alone and either GDC-0941 or GDC-0980 alone. (30B) Immunoblots from three AR-expressing TNBC cell lines treated with either bicalutamide (CDX), GDC-0941, GDC0980 alone or bicalutamide in combination with either GDC-0941 or GDC-0980 for 24h 15 or 48h were probed for AR, PARP, p-AKT, AKT, p-S6, S6 and GAPDH. Figures 31A-31B: Combined inhibition of AR and P13K increases apoptotic cell death in AR+ TNBC cell lines. (3 1A) Bar graphs display relative caspase 3/7 activity (RLU) normalized to viable cell number 48 hrs after treatment with vehicle, positive control (3pM adriamycin, ADR), 50pM bicalutamide (CDX) and 3pM 20 GDC-0941 or I M GDC-0980 alone or in combination with CDX. (3 1B) Histograms show representative cell cycle analysis of MDA-MB-453 cells treated with similar conditions as above. Indicated percentages represent the fraction of hypodiploid DNA (sub-G1), indicative of late stage apoptotic DNA fragmentation. Figures 32A-32C: Simultaneous targeting of AR and P13K decreases 25 viability of AR-expressing cell lines grown in a 3-D forced suspension assay and inhibits in vivo xenograft growth. (3 1A) Representative images display 3-D cell aggregates of MDA-MB-453 cells treated with increasing doses of GDC-0941 or GDC-0980 in the absence or presence of 25pM bicalutamide (CDX). (32B) Line graphs display relative viability of 3-D cell aggregates treated with GDC-0941 or 30 GDC-0980 alone or in combination with 25pM bicalutamide (CDX). Dashed black line depicts the theoretical line of additivity determined from the effect of WO 2013/075059 PCT/US2012/065724 - 12 bicalutamide alone and either GDC-0941 or GDC-0980 alone. (32C) Nude mice bearing established tumors from AR-positive TNBC cell lines (CAL-148 and MDA MB-453) were treated with either bicalutamide (CDX, black hashed line), NVP BEZ235, GDC-0980 or with the combination of bicalutamide and either NVP 5 BEZ235 (blue hashed line) or GDC-0980 (red hashed line). Serial tumor volumes (mm 3 ) were measured at the indicated days. Each data point represents mean tumor volume of 16 tumors; error bars represent SEM. Figures 33A-33B: Detection of PIK3CA mutations by Sanger and digital droplet PCR. (33A) Electropherograms display representative results from sanger 10 sequencing of the E20 amplicon in a PIK3CA wild-type (top) or a H1047R PIK3CA mutant sample. (33B) Scatterplot shows the percent of cells with mutant PIK3CA detected by digital droplet compared to Sanger sequencing. Figure 34: Genetic targeting of AR increases the efficacy of the P13K inhibitor BKM-120 and PI3K/mTOR inhibitor NVP-BEZ235. Line graphs display 15 relative viability of LAR cell lines infected with nontargeting (shNT) or shRNAs targeting AR (shARI and shAR2) following a 72h treatment with the pan-PI3K inhibitor BKM-120 (top) or the dual PI3K/mTOR inhibitor NVP-BEZ235. Figure 35: Pharmacological targeting of AR with bicalutamide increases the efficacy of BKM120 and NVP-BEZ235 in AR-expressing TNBC. Line graphs show 20 relative viability of AR-expressing cell lines treated with increasing doses of BKM120 (top) or NVP-BEZ235 (bottom) alone or in combination with 25pM bicalutamide (CDX). Dashed black line depicts the theoretical line of additivity determined from the effect of bicalutamide alone and either BKM120 or NVP BEZ235 alone. 25 Figures 36A-36B: Simultaneous targeting of AR and P13K decreases viability of AR-expressing cell lines grown in a 3-D forced suspension assay. (36A) Line graphs display relative viability of 3-D cell aggregates treated with BKM120 or NVP-BEZ235 alone or in combination with 25uM bicalutamide (CDX). Dashed black line depicts the theoretical line of additivity determined from the effect of 30 bicalutamide alone and either BKM120 or NVP-BEZ235 alone. (36B) Immunoblots from three AR-expressing TNBC cell lines treated with either bicalutamide (CDX), WO 2013/075059 PCT/US2012/065724 - 13 BKM-120, BEZ235 alone or in combination with bicalutamide for 24h or 48h. Western blots were probed for AR, PARP, p-AKT, AKT, p-S6, S6 and GAPDH. Figure 37: Combined Targeting of P13K and AR increase apoptosis detected by annexin V and propidium iodide (PI) staining. Scattergrams show the percentage 5 of viable (lower left quadrant, annexin V and PI negative), early apoptotic (lower right quadrant, annexin V-positive and PI-negative), or late apoptotic/necrotic (upper right quadrant, annevin V and PI positive) cells following 48 hrs of treatment with either GDC-0941 or GDC-0980 alone or in combination with 50pM bicalutamide (CDX). 10 Figures 38A-38C: Genetic and pharmacological targeting of AR increases p AKT. Figure 39: PDGFR inhibition selectively inhibits viability of mesenchymal like TNBC cell lines. Bar graph displays the half-maximal inhibitory concentration (nM) of the PDGFR inhibitor sorafenib for a panel of TNBC cell lines. Colorbar 15 below designates the TNBC subtype (basal-like= black, mesenchymal-like= grey) or normal cells (white). Figure 40: IGFIR inhibition selectively inhibits viability of mesenchymal like TNBC cell lines. Bar graph displays the half-maximal inhibitory concentration (nM) of the IGF1R inhibitor OSI-906 for a panel of TNBC cell lines. Colorbar 20 below designates the TNBC subtype (basal-like= black, mesenchymal-like=grey) or normal cells (white). Figure 41: IGFlR inhibition selectively inhibits viability of mesenchymal like TNBC cell lines. Bar graph displays the half-maximal inhibitory concentration (nM) of the IGF1R inhibitor BMS-754807 for a panel of TNBC cell lines. Colorbar 25 below designates the TNBC subtype (basal-like= black, mesenchymal-like= grey) or normal cells (white). Figure 42: MET inhibition selectively inhibits viability of mesenchymal-like TNBC cell lines. Bar graph displays the half-maximal inhibitory concentration (nM) of the MET inhibitor PF2341066 for a panel of TNBC cell lines. Colorbar below 30 designates the TNBC subtype (basal-like= black, mesenchymal-like= grey) or normal cells (white).

WO 2013/075059 PCT/US2012/065724 -14 DETAILED DESCRIPTION OF THE INVENTION Triple-negative breast cancer (TNBC) is a highly diverse group of cancers, and subtyping is necessary to better identify molecular-based therapies. In this study, 5 gene expression (GE) profiles from 21 breast cancer data sets were analyzed and 587 TNBC cases were identified. Cluster analysis identified 6 TNBC subtypes displaying unique GE and ontologies, including 2 basal-like (BL and BL2), an immunomodulatory (IM), a mesenchymal (M), a mesenchymal stem-like (MSL), and a luminal androgen receptor (LAR) subtype. Further, GE analysis allowed 10 identification of TNBC cell line models representative of these subtypes. Predicted "driver" signaling pathways were pharmacologically targeted in these cell line models as proof of concept that analysis of distinct GE signatures can inform therapy selection. BL1 and BL2 subtypes had higher expression of cell cycle and DNA damage response genes, and representative cell lines preferentially responded 15 to cisplatin. M and MSL subtypes were enriched in GE for epithelial-mesenchymal transition, and growth factor pathways and cell models responded to NVP-BEZ235 (a PI3K/mTOR inhibitor) and dasatinib (an abl/src inhibitor). The LAR subtype includes patients with decreased relapse-free survival and was characterized by androgen receptor (AR) signaling. LAR cell lines were uniquely sensitive to 20 bicalutamide (an AR antagonist). These data provide biomarker selection, drug discovery, and clinical trial design that align TNBC patients to appropriate targeted therapies. An extensive number of TNBC GE profiles was compiled with the intent of molecularly subtyping the disease. Six (6) TNBC subtypes were identified. Further, 25 using GE signatures derived from each TNBC subtype, representative TNBC cell lines that serve as models for the different subtypes of the disease were aligned. Using the panel of cell lines, prominent signaling pathways revealed by GE signatures were pharmacologically targeted and it was found that the cell lines representing the various subtypes had different sensitivities to targeted therapies 30 currently under laboratory and clinical investigation. The identification of diverse WO 2013/075059 PCT/US2012/065724 - 15 TNBC subtypes and the molecular drivers in corresponding cell line models provides great insight to the heterogeneity of this disease and provides preclinical platforms for the development of effective treatment (see Lehmann, B.D., et al., J Clin Invest, 121(7):2750-2767 (July 2011), the entire teachings of which are 5 incorporated herein by reference). Also described herein is further investigation of a panel of AR-positive TNBC tumors which showed that PIK3CA kinase mutations are a frequent event in the LAR subtype (65.3% vs. 3 8.4%) and were not just selected for in vitro during establishment of the tumor-derived cell lines. Also shown herein is that P13K 10 pathway activation plays a role in resistance to AR antagonists, as there are elevated levels of activated AKT in residual tumors after bicalutamide treatment of xenograft tumors. Further, it was found that genetic or pharmacological targeting of AR synergizes with PI3K/mTOR inhibition in both two- and three-dimensional cell culture models. Based on the cell culture-based results, the combination of these 15 inhibitors in in vivo xenograft studies was examined using bicalutamide +/ GDC0980 or NVP-BEZ235. The preclinical data herein provide the rationale for pre-selecting AR-positive TNBC patients for treatment with the combination of AR antagonists and/or PI3K/mTOR inhibitors. As discussed herein, triple negative breast cancer (TNBC) refers to breast 20 cancers that are estrogen receptor (ER) negative, progesterone receptor (PR) negative and human epidermal growth factor receptor 2 (HER-2) negative. The invention is based, in part, of the discovery of six TNBC subtypes. Accordingly, in one aspect provided herein are methods of determining a triple negative breast cancer (TNBC) subtype in an individual in need thereof 25 comprising determining expression of one or more genes (e.g., presence of one or more mRNAs and/or protein encoded by gene) in one or more TNBC cells of the individual; and comparing the expression of the one or more genes in the TNBC cells with the expression of the one or more genes in a control. Increased expression of one or more genes comprising one or more cell cycle genes, cell division genes, 30 proliferation genes, DNA damage response genes or a combination thereof in the TNBC cells compared to the control indicates that the TNBC is a TNBC BLI WO 2013/075059 PCT/US2012/065724 - 16 subtype. Increased expression of one or more growth factor signaling genes, glycolysis genes, gluconeogenesis genes or a combination thereof in the TNBC cells compared to the control indicates that the TNBC is a TNBC BL2 subtype. Increased expression of one or more immune cell signaling genes, cytokine signaling genes, 5 antigen processing and presentation genes, core immune signal transduction genes or a combination thereof in the TNBC cells compared to the control indicates that the TNBC is a TNBC IM subtype. Increased expression of one or more cell motility genes, extracellular matrix (ECM) receptor interaction genes, cell differentiation genes or a combination thereof in the TNBC cells compared to the control indicates 10 that the TNBC is a TNBC M subtype. Increased expression of one or more cell motility genes, cellular differentiation genes, growth pathway genes, growth factor signaling genes, angiogenesis genes, immune signaling genes, stem cell genes, HOX genes, mesenchymal stem cell-specific marker genes or a combination thereof in the TNBC cells compared to the control indicates that the TNBC is a TNBC MSL 15 subtype. Increased expression of one or more hormone regulated genes in the TNBC cells compared to the control indicates that the TNBC is a TNBC LAR type. In a particular aspect, the invention is directed to a method of determining a TNBC BLI subtype in an individual in need thereof comprising determining increased expression of one or more cell cycle genes, cell division genes, 20 proliferation genes, DNA damage response genes or a combination thereof in the TNBC cells of the individual compared to the control. Increased expression of the one or more genes indicates that the TNBC is a BLI subtype TNBC. In one aspect, the method of determining a TNBC BLI subtype in an individual in need thereof comprises selectively detecting increased expression of a gene combination 25 comprising AURKA, AURKB, CENPA, CENPF, BUB1, BUB1B, TTK, CCNA2, PLKJ, PRCl, MYC, NRAS, PLKJ, BIRC5, CHEK], FANCA, FANCG, FOXM1, HNGA1, RAD54BP, RAD51, NEKS, NBN, EXO1, MSH2, MCM1O, RAD21,SIX3, ZiC1, SOX4, SOX1 and MDC1 wherein detection of increased expression of the gene combination indicates that the TNBC is TNBC BLI subtype. Determination 30 that the TNBC is a TNBC BLI subtype can further comprise detecting increased expression of Ki-67 mRNA in the TNBC cells compared to the control. In addition, WO 2013/075059 PCT/US2012/065724 -17 determination that the TNBC is a TNBC BLI subtype can further comprise detecting one or more mutated genes. Examples of such mutated genes include mutated BRCA1, STAT4, UTX, BRCA2, TP53, CTNND1, TOP2B, CAMKlG, MAPK13, MDC1, PTEN, RB1, SMAD4, CDKN2A, ATM, ATR, CLSPN, HDAC4, 5 NOTCH, SMARCAL1, and TIMELESS. In another aspect, the invention is directed to a method of determining a TNBC BL2 subtype in an individual in need thereof comprising determining increased expression of one or more growth factor signaling genes, glycolysis genes, gluconeogenesis genes or a combination thereof in the TNBC cells compared to the 10 control. Increased expression of the one or more genes indicates that the TNBC is a TNBC BL2 subtype. In one aspect, the method of determining a TNBC BL subtype in an individual in need thereof comprises selectively detecting increased expression of a gene combination comprising EGFR, MET, ELF4, MAF, NUAKI, JAG1, FOSL2, IDI, ZICI, SOXI1, 1D3, FHL2, EPHA2, TP63 and MME wherein 15 detection of increased expression of the gene combination indicates that the TNBC is TNBC BL2 subtype. Determination that the TNBC is a TNBC BL2 subtype can further comprise detecting one or more mutated genes. Examples of such genes comprise mutated BRCA1, RB1, TP53, PTEN, CDKN2A, UTX, BRAC2, PTCH1, PTCH2, and RET. 20 In another aspect, the invention is directed to a method of determining a TNBC M subtype in an individual in need thereof comprising determining increased expression of one or more cell motility genes, extracellular matrix (ECM) receptor interaction genes, cell differentiation genes or a combination thereof in the TNBC cells compared to the control. Increased expression of the one or more genes 25 indicates that the TNBC is a TNBC M subtype. In one aspect, the method of determining a TNBC M subtype in an individual in need thereof comprises selectively detecting increased expression of a gene combination comprising TGFBiL1, BGN, SMAD6, SMAD7, NOTCH, TGFB1, TGFB2, TGFB3, TGFBR1, TGFBR2, TGFBR3, MMP2, ACTA2, SNAI2, SPARC, SMAD7, 30 PDGFRA, TAGLN, TCF4, TWIST, ZEB1, COL3A1, JAG1, EN1, MYLK, STK38L, CDH11, ETV5, IGF1R, FGFR1, FGFR2, FGFR3, TBX3, COL5A2, WO 2013/075059 PCT/US2012/065724 - 18 GNG11, ZEB2, CTNNB1, DKK2, DKK3, SFPR4, TCF4, TCF7L2, FZD4, CAV1, CAV2, CCND 1 and CCND2 wherein detection of increased expression of the gene combination indicates that the TNBC is TNBC M subtype. Determination that the TNBC is a TNBC M subtype can further comprise detecting decreased E-cadherin 5 (CDH1) expression compared to the control. Determination that the TNBC is a TNBC M subtype can further comprise detecting one or more mutated genes. Examples of such mutated genes include mutated PTEN, RB1, TP53, PIK3CA, APC, BRAF, CTNNB1, FGFR1, GLIl, HRAS, KRAS, NOTCH, and NOTCH4, Determination that the TNBC is a TNBC M subtype can further comprise detecting 10 chromosomal amplifications in KRAS, IGFIR or MET. In another aspect, the invention is directed to a method of determining a TNBC MSL subtype in an individual in need thereof comprising determining increased expression of one or more cell motility genes, cellular differentiation genes, growth pathway genes, growth factor signaling genes, angiogenesis genes, 15 immune signaling genes, stem cell genes, HOX genes, mesenchymal stem cell specific marker genes or a combination thereof in the TNBC cells compared to the control. Increased expression of the one or more genes indicates that the TNBC is a TNBC MSL subtype. In one aspect, the method of determining a TNBC MSL subtype in an individual in need thereof comprises selectively detecting increased 20 expression of a gene combination comprising VEGFR2 (KDR), TEK, TIEl, EPASI, ABCA8, PROCR, ENG, ALDHA1, PER1, ABCB1, TERF2IP, BCL2, BMP2, EPASI, STAT4, PPARG, JAKI, IDI, SMAD3, TWIST, THY1, HOXA5, HOXA1O, GL13, HHEX, ZFHX4, HMBOXI, FOS, PIK3R1, MAF, MAFB, RPS6KA2, TCF4, TGFB1L1, MEIS1, MEIS2, MEOXI, MEOX2, MSX1, ITGAV, 25 KDR, NGFR, NT5E, PDGFRA, PDGFRB, POU2F1, and VCAM1 wherein detection of increased expression of the gene combination indicates that the TNBC is TNBC MSL subtype. Determination that the TNBC is a TNBC MSL subtype can further comprise detecting low expression of claudins 3, 4, 7 or a combination thereof compared to the control. Determination that the TNBC is a MSL subtype 30 TNBC can further comprise detecting one or more mutated genes. Examples of such WO 2013/075059 PCT/US2012/065724 - 19 mutated genes include mutated CDKN2A, HRAS, TP53, NFl, PIK3CA, BRCA1, BRAF, KRAS, NF2, PDGFRA, APC, CTNNB1, FGFR1, PDGFRB. In another aspect, the invention is directed to a method of determining a TNBC IM subtype in an individual in need thereof comprising determining 5 increased expression of one or more immune cell signaling genes, cytokine signaling genes, antigen processing and presentation genes, core immune signal transduction genes or a combination thereof in the TNBC cells compared to the control. Increased expression of the one or more genes indicates that the TNBC is a TNBC IM subtype. In one aspect, the method of determining a TNBC IM subtype in an individual in 10 need thereof comprises selectively detecting increased expression of a gene combination comprising CCL19, CCL3, CCL4, CCL5, CCL8, CCR1, CCR2, CCR5, CCR7, CD2, CD37, CD38, CD3D, CD48, CD52, CD69, CD74, and CD8A wherein detection of increased expression of the gene combination indicates that the TNBC is TNBC IM subtype. Determination that the TNBC is a TNBC IM subtype can 15 . further comprise detecting one or more mutated genes. Examples of such mutated genes include TP53, CTNNA1, DDX18, HUWEl, NFKBIA, APC, BRAF, MAP2K4, RB1, STAT4, STAT1, and RET. In another aspect, the invention is directed to a method of determining a TNBC LAR subtype in an individual in need thereof comprising determining 20 increased expression of one or more hormone regulated genes in the TNBC cells compared to the control. Increased expression of the one or more genes indicates that the TNBC is a TNBC LAR type. In one aspect, the method of determining a TNBC LAR subtype in an individual in need thereof comprises selectively detecting increased expression of a gene combination comprising AR, DHCR24, ALCAM, 25 GATA2, GATA3, IDIHI, IDIH2, CDH1 1, ERBB3, CUX2, FGFR4, HOPX, FASN, FKBP5, APOD, PIP, SPDEF, CLDN8, FOXA1, KRT18,and XBP1 wherein detection of increased expression of the gene combination indicates that the TNBC is TNBC LAR subtype. Determination that the TNBC is a LAR subtype TNBC can further comprise detecting one or more mutated genes. Examples of such mutated 30 genes include PIK3CA, CDHI, PTEN, RB1, TP53, and MAP3K1.

WO 2013/075059 PCT/US2012/065724 - 20 A variety of methods (e.g. spectroscopy, colorometry, electrophoresis, chromatography) can be used to detect increased expression of a (one or more) gene. As is apparent to those of skill in the art, increased expression of one or more genes can be detected by measuring all or a portion of nucleic acid (e.g., DNA, mRNA) of 5 the gene and/or protein (polypeptide) expressed by the gene. The level, expression and/or activity of a gene and or its encoded polypeptide can be measured. For example, nucleic acid such as genes, mRNA or a combination thereof, can be determined using PCR (e.g., RT-PCR, RT-qPCR), gene chips or microarray analysis (DNA arrays, gene expression arrays, nanostrings), next generation sequencing (e.g., 10 next generation RNA sequencing), in situ hybridization, blotting techniques, and the like. Protein can be detected using immunoassays (e.g.,immunohistochemistry, immunofluorescence, immunoprecipitation), arrays (e.g., reverse phase protein microarrays), blotting techniques (e.g., Western blots), SDS-PAGE, and the like As described herein, the methods can further comprise detecting one or more 15 mutated genes. A variety of methods can be used to detect one or more mutated genes. Examples of such methods include a SNaPshot assay, exome sequencing, sanger gene sequencing, resequencing array analysis, mRNA analysis/cDNA sequencing polymerase chain reaction (PCR), single-strand conformation polymorphism (sscp), heteroduplex analysis (het), allele-specific oligonucleotide 20 (aso), restriction fragment analysis, allele-specific amplification (asa), single nucleotide primer extension, oligonucleotide ligation assay (ola), denaturing gradient gel electrophoresis (DGGE), temperature gradient gel electrophoresis (TGGE) and single strand conformation polymorphism (SSCP). The methods can further comprise comparing the expression of the one or 25 more genes in the individual to the expression of the one or more genes in a control. As will be apparent to those of skill in the art, a variety of suitable controls can be used. Examples of such controls include one or more non-cancer cells (e.g., breast cells from one or more individuals that do not have cancer) and/or non-TNBC cells (e.g., from the individual being tested). The control can also be a (one or more) 30 standard or reference control established by assaying one or more (e.g., a large sample of) individuals which do not have cancer (e.g., TNBC) and using a statistical WO 2013/075059 PCT/US2012/065724 -21 model to obtain a control value (standard value; known standard). See, for example, models described in Knapp, R.G. and Miller M.C. (1992) Clinical Epidemiology and Biostatistics, William and Wilkins, Harual Publishing Co. Malvern, PA, which is incorporated herein by reference. 5 As used herein an "individual" refers to an animal, and in a particular aspect, a mammal. Examples of mammals include primates, a canine, a feline, a rodent, and the like. Specific examples include humans, dogs, cats, horses, cows, sheep, goats, rabbits, guinea pigs, rats and mice. The term "individual in need thereof' refers to an individual who is in need 10 of treatment or prophylaxis as determined by a researcher, veterinarian, medical doctor or other clinician. In one aspect, an individual in need thereof is a mammal, such as a human. In other aspects, the individual has been diagnosed with breast cancer, the individual has not been diagnosed with breast cancer, or the individual is at risk of developing breast cancer. 15 The methods of the invention can further comprise detecting expression (increased expression) of one or more genes in a sample (biological sample) of the individual. Thus, the method can further comprise obtaining a (one or more) sample from the individual. The sample can be, for example, a biological fluid, a biological tissue (e.g., a biopsy) and combinations thereof from the individual. In particular 20 aspects, the sample is one or more TNBC cells (e.g., epithelial cells, mesenchymal cells, immune cells). Methods for obtaining such biological samples from an individual are known to those of skill in the art. As also shown herein, the methods can further comprise determining a treatment protocol for the triple negative breast cancer (TNBC) patient based on the 25 TNBC subtype. As described herein, a "cancer therapy", "treatment for cancer" or "cancer treatment" comprises administering one or more agents to the individual. In one embodiment, the agent is a chemotherapeutic agent which targets the TNBC cells. Examples of such agents include an alkylating agent (e.g., busulfan, cisplatin, carboplatin, chlorambucil, cyclophosphamide, ifosfamide, dacarbazine (DTIC), 30 mechlorethamine (nitrogen mustard), melphalan, temozolomide); a nitrosoureas (e.g., carmustine (BCNU), lomustine (CCNU)); an antimetabolite (e.g., 5- WO 2013/075059 PCT/US2012/065724 - 22 fluorouracil, capecitabine, 6-mercaptopurine, methotrexate, gemcitabine, cytarabine (ara-C), fludarabine, pemetrexed); an anthracycline or related drug (e.g., daunorubicin, doxorubicin (adriamycin), epirubicin, idarubicin, mitoxatrone); a topoisomerase (topoisomerase I inhibitor; topoisomerase inhibitor II) inhibitor (e.g., 5 topotecan, irinotecan, etoposide (VP-16), teniposide); and a mitotic inhibitor (e.g., taxanes (paclitaxel, docetaxel, taxol) vinca alkaloids (vinblastine, vincristine, vinorelbine). In another embodiment, the agent is a targeted therapy agent which attacks cancer cells more specifically than chemotherapeutic agents. Examples of such 10 agents include imatinib (Gleevec), gefitinib (Iressa), erlotinib (Tarceva), rituximab (Rituxan), and bevacizumab (Avastin). In yet another embodiment, the agent is a sex hormone, or hormone-like drug that alters the action or production of female or male hormones and can be used to slow the growth of breast, prostate, and endometrial (uterine) cancers. Examples of such agents include anti-estrogens (e.g., tamoxifen, 15 fulvestrant), aromatase inhibitors (e.g., anastrozole, exemestane, letrozole), progestins (e.g., megestrol acetate), anti-androgens (e.g., bicalutamide, flutamide) and LHRH agonists (leuprolide, goserelin). In addition, the agent can be a drug which is used to stimulate the immune system to more effectively recognize and attack cancer cells of an individual with cancer. Other examples of agents include a 20 hormone such as a selective estrogen receptor modulator (SERM); an antibody or antigen binding fragment thereof (e.g., herceptin); a protein tyrosine kinase inhibitor; a combination of chemotherapeutic agents such as cytoxan (C), methotrexate (M), fluorouracil (F), an anthracylcin such as adriamycin (A), epirubicin (E), doxorubicin, and/or daunorubicin; a targeted kinase inhibitor; a 25 metalloproteinase inhibitor; and a proteosome inhibitor. The one or more agents can be administered to the individual before (e.g., neoadjuvant therapy), during or after (e.g., adjuvant therapy) primary treatment of the cancer. As used herein "primary treatment of a cancer" generally refers to a treatment involving surgery (e.g., surgical removal of a tumor) and/or radiation (e.g., 30 local radiation). In the methods of the present invention, the "treatment for cancer" can be a neoadjuvant treatment wherein an agent (e.g., a hormone, a WO 2013/075059 PCT/US2012/065724 - 23 chemotherapeutic) is administered prior to surgical removal of a (residual) tumor, or prior to localized radiation (e.g., radiation used to shrink the tumor in order to simplify subsequent surgical removal of the tumor). In another embodiment, the "treatment for cancer" can be an adjuvant treatment wherein an agent (e.g., a 5 hormone, a chemotherapeutic agent) is administered after surgical removal of a (residual) tumor, or after localized radiation. In yet another embodiment, the treatment can be a palliative treatment in which one or more chemotherapeutic agents are administered to an individual to treat metastatic cancer (e.g., to make the individual more comfortable and/or to prolong survival of the individual after the 10 cancer has metastasized). In particular aspects, the TNBC BLI or BL2 subtype is detected in the individual and the individual is treated with one or more drugs that damages DNA. Examples of such drugs include an alkylating-like agent (e.g., cisplatin) and a PARP inhibitor (e.g., veliparib, olaparib). 15 In another aspect, the TNBC ML subtype is detected in the individual and the individual is treated with one or more drugs that inhibits Src (e.g., disatinib), IGF1R (e.g., OSI-906, BMS-754807), MET (e.g., PF2341066), PDGFR (e.g., sorafinib), P13K (e.g., BKM-120, GDC0941), Pl3K/mTOR (e.g., NVP-BEZ235, GDC0980) or a combination thereof. 20 In another aspect, the TNBC LAR subtype is detected in the individual and the individual is treated with one or more drugs that inhibit androgen receptor (AR). Examples of such drugs include bicalutamide, MVD3 100, and abiraterone. The method can further comprise treating the individual with P13K/mTOR (e.g., NVP BEZ235, GDC0980) or a combination thereof. The method can further comprise 25 treating the individual with an inhibitor of HSP90 (e.g., DMAG). In other aspects, the invention can further comprise predicting whether an individual (e.g., patient) will benefit from a (one or more) treatment for a particular TNBC subtype. Also described herein are TNBC cell lines that serve as models for each of 30 the six TNBC subtypes. Thus, in another aspect, the invention is directed to a method of determining whether an agent can be used to treat a triple negative breast WO 2013/075059 PCT/US2012/065724 - 24 cancer (TNBC) subtype comprising contacting a cell line that is a model for the TNBC subtype of interest with an agent to be assessed; and determining viability of the cell line in the presence of the agent, wherein if the viability of the cell line decreases in the presence of the agent, then the agent can be used to treat the TNBC 5 subtype of interest. The TNBC subtype of interest can be a TNBC BL-1 subtype, a TNBC BL-2 subtype, a TNBC IM subtype, a TNBC M subtype, a TNBC MSL subtype or a TNBC LAR subtype. In addition, as will be apparent to those of skill in the art, the viability of a cell decreases when e.g., the cell undergoes apoptosis, proliferation of the cell is inhibited or slowed and/or metastasis of the cell is 10 inhibited to slowed. Examples of agents (test agent; agent of interest) include nucleic acids, polypeptides, fusion proteins, peptidomimetics, prodrugs, drugs, receptors, binding agents (e.g., antibodies), small molecules, etc and libraries (e.g., combinatorial libraries) of such agents. Test agents can be obtained made or obtained from 15 libraries of natural, synthetic, and/or semi-synthetic products (e.g., extracts). Those skilled in the field of drug discovery and development will understand the precise source of such agents. In one aspect, the invention is directed to a method of determining whether an agent can be used to treat a triple negative breast cancer (TNBC) BL-1 subtype 20 comprising contacting one or more TNBC BL-1 subtype cell lines with an agent to be assessed. The viability of the one or more cell lines in the presence of the agent is detected, wherein if the viability of the one or more cell lines decreases in the presence of the agent, then the agent can be used to treat the TNBC BL-1 subtype. Examples of TNBC BL-1 subtype cell lines include HCC2157, HCC1599, 25 HCC1937, HCC1 143, HCC3153, MDA-MB-468, and HCC38. In another aspect, the invention is directed to a method of determining whether an agent can be used to treat a triple negative breast cancer (TNBC) BL-2 subtype comprising contacting one or more TNBC BL-2 subtype cell lines with an agent to be assessed. The viability of the one or more cell lines in the presence of the 30 agent is determined, wherein if the viability of the one or more cell lines decreases in the presence of the agent, then the agent can be used to treat the TNBC BL-2 WO 2013/075059 PCT/US2012/065724 -25 subtype. Examples of TNBC BL-2 subtype cell lines include SUM149PT, CAL-851, HCC70, HCC1806 and HDQ-P1. In another aspect, the invention is directed to a method of determining whether an agent can be used to treat a triple negative breast cancer (TNBC) IM 5 subtype comprising contacting one or more TNBC IM subtype cell lines with an agent to be assessed. The viability of the one or more cell lines in the presence of the agent is determined, wherein if the viability of the one or more cell lines decreases in the presence of the agent, then the agent can be used to treat the TNBC IM subtype. Examples of TNBC IM subtype cell lines include HCCI 187 and DU4475. 10 In another aspect, the invention is directed to a method of determining whether an agent can be used to treat a triple negative breast cancer (TNBC) M subtype comprising contacting one or more TNBC M subtype cell lines with an agent to be assessed. The viability of the one or more cell lines in the presence of the agent is determeined, wherein if the viability of the one or more cell lines decreases 15 in the presence of the agent, then the agent can be used to treat the TNBC M subtype. Examples of TNBC M subtype cell lines include BT-549, CAL-51 and CAL-120. In another aspect, the invention is directed to a method of determining whether an agent can be used to treat a triple negative breast cancer (TNBC) MSL 20 subtype comprising contacting one or more TNBC MSL subtype cell lines with an agent to be assessed. The viability of the one or more cell lines in the presence of the agent is determined, wherein if the viability of the one or more cell lines decreases in the presence of the agent, then the agent can be used to treat the TNBC MSL subtype. Examples of TNBC MSL subtype cell lines include HS578T, MDA-MB 25 157, SUM159PT, MDA-MB-436, and MDA-MB-231. In another aspect, the invention is directed to a method of determining whether an agent can be used to treat a triple negative breast cancer (TNBC) LAR subtype comprising contacting one or more TNBC LAR subtype cell lines with an agent to be assessed. The viability of the one or more cell lines in the presence of the 30 agent is determined, wherein if the viability of the one or more cell lines decreases in the presence of the agent, then the agent can be used to treat the TNBC LAR WO 2013/075059 PCT/US2012/065724 - 26 subtype. Examples of TNBC LAR subtype cell lines inclulde MDA-MB, SUM185PE, HCC2185, CAL-148, and MFM-223. In the methods of the invention, viability of one of more of the cells can be determined using a variety of methods. Examples of such methods include a 5 membrane leakage assay, a mitochondrial activity assay, a functional assay, a proteomic assay, a genomic assay or a combination thereof. The methods can further comprise comparing the viability of the one or more cell lines in the presence of the agent to the viability of a control such as a non cancerous cell line (e.g., a culture of normal human and immortalized MCF210A). 10 Exemplification Example I Identification of human triple negative breast cancer subtypes and models for selection of targeted therapies Methods 15 Laser capture microdissection, RNA extraction, and GE profiling. Invasive tumor cells from serial sections of primary breast cancers (n = 112) were captured onto polymeric caps using the PixCell II lasercapture microdissection system (Arcturus) as previously described (Bauer JA, et al. Clin Cancer Res. 2010; 16(2):681-690). Areas of ductal carcinoma in situ and normal breast tissue were 20 excluded and areas of inflammation and areas with tumor-associated fibroblasts were avoided. Total RNA was isolated from captured cells, quantified, and integrity analyzed, and microarray analyses were done using Affymetrix GeneChip Human Genome U133 Plus 2.0 arrays as previously described (Bauer JA, et al. Clin Cancer Res. 2010; 16(2):681-690). 25 Data set collection and TNBC identification by bimodal filtering. 2353 human breast cancer GE profiles were compiled from 14 publicly available breast cancer microarray data sets (GEO, www.ncbi.nlm.nih.gov/gds; Array Express, www.ebi.ac.uk/microarray-as/ae/) including 112 primary breast cancer GE profiles from our institution (Vanderbilt University Medical Center). All Vanderbilt tissue 30 samples were taken from individuals treated at Vanderbilt University with institutional review board approval, and all patients signed a protocol-specific WO 2013/075059 PCT/US2012/065724 -27 consent. All GE profiles were generated on Affymetrix microarrays and collected for identification of TNBCs for the training data set (Table 1). An additional 894 breast cancer GE profiles from 7 data sets were collected for the identification of TNBCs for the validation data set (Table 2). Raw GE values for each data set (n = 5 21) were normalized independently using RMA procedure. The Affymetrix probes 205225_at, 208305_at and 216836_s_at were chosen to represent ER, PR, and HER2 expression, respectively (Karn T, et al. Breast Cancer Res Treat. 2010;120(3):567-579). For each data set, empirical expression distributions of ER, PR, and HER2 were analyzed using a 2-component Gaussian mixture distribution 10 model and parameters were estimated by maximum likelihood optimization, using optim function (R statistical software, www.r-project.org). The posterior probability of negative expression status for ER, PR, and HER2 was then estimated. A sample was classified as having negative expression if its posterior probability was less than 0.5. Upon initial hierarchical clustering, it was evident that some samples were from 15 the same tumors, but in different data sets. These and additional outliers were removed after PCA. To ensure all ER/PR/HER2-positive tumors were removed, a secondary filter was performed on the combined samples identified by bimodal filtering. TNBC tumors were renormalized along with 5 positive controls for each parameter (ER, PR, and HER2) from 4 data sets that were positive by IHC and 20 expressed mRNA near the center of the positive bimodal peak (Figure IB). Only tumors that displayed a greater than 10-fold reduction in expression than the positive controls were considered negative in expression and used for further analysis, resulting in 386 TNBC tumors (training set) and 201 TNBC tumors (validation set). Normalization, data reduction, and cross-platform batch effect removal. 25 Training (n = 386) and validation (n = 201) TNBC data sets identified by the 2 component Gaussian distribution model were collectively RMA normalized, summarized and log-transformed using the combined raw GE from each data set. For genes containing multiple probes, the probe with the largest interquartile range across the samples was chosen to represent the gene. Batch effects were removed by 30 fitting each gene to a linear model with 14 and 7 fixed effects for each data set, WO 2013/075059 PCT/US2012/065724 -28 respectively. The residual genes from this model (n = 13,060) were used for subsequent clustering analysis. GE-based identification of TNBC subtypes. k-means clustering and consensus clustering were used to determine the optimal number of stable TNBC 5 subtypes. Cluster robustness was assessed by consensus clustering using agglomerative k-means clustering (1,000 iterations), with average linkage on the 386 TNBC profiles using the most differentially expressed genes (SD > 0.8; n = 1510 genes) (Gene Pattern version 3.2.1, GSE-A, www. broadinstitute.org/gsea/) (19). The optimal number of clusters was determined from the CDF, which plots the 10 corresponding empirical cumulative distribution, defined over the range [0,1], and from calculation of the proportion increase in the area under the CDF curve. The number of clusters is decided when any further increase in cluster number (k) does not lead to a corresponding marked increase in the CDF area. PCA and heat maps were generated using Genespring GX ver. 10 software (Agilent). 15 Functional annotation. Each TNBC subtype was tested for gene enrichment compared with all other samples using GSE-A software (Subramanian A, et al. Proc Natl Acad Sci U S A. 2005;102(43):15545-15550). Genes were tested for enrichment in the C2 curated gene sets of canonical pathways. Using the GSE-A algorithm (1,000 permutations), the top significantly enriched canonical pathways 20 were selected based on a normalized enrichment score (NES) greater than 0.4 and false discovery rate (FDR) q value of less than 0.60. The FDR cutoff of 0.60 was chosen because of the lack of coherence in the data set collection spanning 21 studies, and more stringent FDR cutoffs resulted in fewer results, potentially overlooking significant pathways. 25 TNBC subtype gene signature derivation. The 20% of genes with the highest and lowest expression levels in at least 50% of the samples for each subtype were chosen for further analysis. Within each cluster, for each selected gene, the nonparametric Kruskal-Wallis test was applied to identify genes significantly different from the median GE among all 6 groups. Significant genes with a 30 Bonferroni's adjusted P value of less than 0.05 were included in the combined gene signature (n = 2188) for the TNBC subtypes and used to predict an independent WO 2013/075059 PCT/US2012/065724 - 29 validation set as well as to classify TNBC cell lines. Each sample from the validation set was assigned to a TNBC subtype based on highest Pearson correlation and lowest P value to one of the subtypes derived from the training data set. Samples with correlations differing by a P value of less than 0.05 were considered 5 unclassifiable. TNBC cell-line classification. GE data for TNBC cell lines (GSE-10890 and E-TABM- 157) were correlated (Spearman) to the centroids of the GE signatures for each TNBC subtype. GE data from both the TNBC tumors and cell lines were combined so that each gene was standardized to have mean = 0 and SD = 1. GE 10 profiles from the cell lines were correlated to the centroids for each of the 6 TNBC subtypes. To remove size effects of the 6 gene signatures, the empirical P values were estimated for the correlations using a resampling (bootstrap, n = 1000) approach and estimating correlation coefficients for each resample. Cell lines were assigned to the TNBC subtype with the highest correlation (Table 3), and those that 15 had low correlations (<0.1) or were similar between multiple subtypes (P > 0.05) were considered unclassified. Cell proliferation/viability assays and IC50 determinations. All cell lines and culture conditions used can be found in Supplemental Methods (Table 5). Short tandem repeat (STR) DNA fingerprinting analysis was performed on all TNBC cell 20 lines used in this study (Cell Line Genetics). All cell lines matched STR databases (ATCC, DSMZ, and Asterand), and their identity was confirmed by a certified process at Cell Line Genetics. Breast cancer cell lines and HMECs were seeded (3,000-10,000 cells) in quadruplicate wells in 96-well plates. Cells were incubated in the presence of alamar-Blue (Invitrogen), and baseline (predrug) fluorescence 25 (Ex/Em: 560/590 nm) measurements were obtained following overnight attachment. Medium was then replaced with either fresh medium (control) or medium containing half-log serial dilutions of the following drugs: 0.3-30 pLM olaparib (ChemieTek), 0.3-30 tM veliparib (ChemieTek), 1-100 tM bicalutamide (IPR Pharmaceutical), 0.3-30 pM cisplatin (APP Pharmaceutical), 3-300 nM NVP-BEZ235 (Chemdea), 30 10-1000 nM 17-DMAG (ChemieTek), and 0.1-10 pM dasatinib (LC Laboratories). Viability was determined from measuring fluorescent intensity after metabolic WO 2013/075059 PCT/US2012/065724 -30 reduction of alamarBlue in the presence/absence of the drug after 72 hours. Viability assays were performed in triplicate, and replicates were normalized to untreated wells. Inhibitory concentration (IC 50 ) values were determined after double-log transformation of dose response curves as previously described (Bauer JA, et al. 5 Breast Cancer Res. 2010;12(3):R41). For analysis of cell-line drug assays, data generated from the different cell lines representative of TNBC subtypes were compared using the nonparametric Mann-Whitney U test. All analyses and graphic representations were performed using Prism software (version 4.0c; GraphPad Software), and values are represented as the mean ± SEM. 10 Kaplan-Meier survival analysis and Hr. Log-rank test was used to determine survival significance in TNBC subtypes from Kaplan-Meier survival curves, RFS, and DMFS. Cox proportional hazards model was used to calculate Hr, demonstrating differences in survival by pairwise comparison between TNBC subtypes (P < 0.05). 15 Xenograft tumor studies. Five-week-old female athymic nude-Foxnlnu mice (Harlan Sprague-Dawley) were injected (s.c.) with either approximately 5 x 106 (CAL-51, HCC1806, MDA-MB-468, and SUM185PE) or approximately 10 x 106 (CAL-148 and SUM159PT) cells suspended in medium (200 pl) into each flank using a 22-gauge needle. The protocols describing the process of xenograft tumor 20 formation in athymic mice were reviewed and approved by the Vanderbilt Institutional Review Board. Once tumors reached a volume of 25-50 mm 3 , mice were randomly allocated to treatment or vehicle arms. Treatments included bicalutamide per oral (100 mg/kg/d), cisplatin i.p. (8 mg/kg/wk), or NVP-BEZ235 per oral (50 mg/kg/d) in 1:9 N-methylpyrolidone: PEG300. Tumor diameters were 25 serially measured at the indicated times with digital calipers, and tumor volumes were calculated as width 2 x length/2. Data points reflect the means and SEM for 16 tumors per treatment. Statistics. Description of the relevant statistical methods used and analyses performed are described in the relevant Methods sections above. Comparisons 30 between cell lines representative of TNBC subtypes were performed using the WO 2013/075059 PCT/US2012/065724 -31 nonparametric Mann-Whitney U test. Statistical significance of drug effects on tumor growth in athymic mice was determined by 2-tailed unpaired t test. Supplemental Methods 5 Immunostaining. Both formalin fixed paraffin embedded (FFPE) and frozen tissue were used for immunohistochemical studies. When frozen tissue was used for immunohistochemistry, staining was performed on sections taken from the same tissue block from which RNA was isolated for microarray. AR, Ki67, EGFR and CK 5/6 expression were evaluated in frozen tissue using the DAKO (Carpinteria, CA) 10 antibodies: AR (clone AR41 1) at a 1:50 dilution, EGFR at 1:200, CK5/6 at 1:50 and Ki67 antibody (MIB 1 clone) at a 1:200 dilution for I h at room temperature. FFPE tissue was subject to antigen retrieval with high pH buffer (pH 8.0) followed by overnight incubation with an AR (1:30) or Ki67 (1:75) antibody dilution overnight. AR expression was scored as both the percentage of tumor cells with nuclear 15 staining as well as the intensity of staining (scored as 0-3+). An AR intensity score was calculated as follows: (AR intensity x 100) + % AR positive cells. For Ki67 the percentage of cells demonstrating nuclear staining at any intensity was recorded. Immunoblotting. Cells were trypsinized, lysed and relative protein expression was determined by Western blot as previously described with the 20 following antibodies; HSP90 monoclonal antibody, clone F-18 (Santa Cruz Biotechnology, Santa Cruz, CA) and the AR polyclonal antibody, SC-N20 (Santa Cruz Biotechnology). Colony Formation. MFM-223, MDA-MB-453 and CAL-148 cells (3000 cells/well) were reverse-transfected with 1.25 pmole of siRNAs to AR 25 [ONTARGET plus SMARTpool, cat# L-003400-00 (Dharmacon, Lafayette, CO)] or non-targeting control (ON-TARGETplus Non-targeting Pool cat# D-001810-10-05) with 0.25 pL Dharmafect #3 (MFM-223), or 0.25 pL Dharmafect #1 (MDAMB-453 and CAL-148). Colonies were stained and quantified 14 d following transfection using Cell Profiler 2.0 (Broad Institute, Cambridge, MA). Experiments were 30 preformed in triplicate and error bars reflect standard deviation.

WO 2013/075059 PCT/US2012/065724 - 32 Results Analysis of human tumor GE profiles identifies TNBC subtypes. GE profiles were obtained from 21 publicly available data sets that contained 3,247 primary human breast cancers and processed according to the flow chart in Figure 1A. To 5 allow for robust analysis, these data sets were further divided into training (Table 1; 14 data sets, cases n = 2353) and validation data sets (Table 2; 7 data sets, cases n = 894). Since the majority of these tumors lacked sufficient molecular analysis of ER, PR, and HER2, we filtered each data set for ER, PR, and HER2 mRNA expression to identify triple-negative status (see Methods for bimodal filter description; 10 supplemental material available in Lehmann, B.D., et al., J Clin Invest, 121(7):2750-2767 (July 2011), the entire teachings of which are incorporated herein by reference). Previous studies have shown that ER and HER2 mRNA expression correlates with immunohistochemistry (IHC) and FISH analyses, respectively (Carmeci C. et al. Am J Pathol. 1997;150(5):1563-1570; Press MF, et al. Clin 15 Cancer Res. 2008;14(23):7861-7870). Using a bimodal filter on the GE data (Figure 1B), 386 and 201 TNBC tumor samples were identified in the training and validation sets, respectively. The 386 TNBC GE profiles of the training set were robust multiarray average (RMA) normalized, summarized, transformed, and corrected for "batch effect," resulting in 13,060 identical probes representing unique 20 genes across all platforms. The triple-negative GE pattern is shown compared with 5 positive controls for each parameter (ER, PR, and HER2) from data sets that were confirmed IHC positive and expressed mRNA near the center of the positive bimodal peak (Figure 1C). Of the 14 data sets in the training set, 4 (GSE-7904, E TABM-158, MDA133, GSE-22513, GSE-28821, and GSE-28796) included IHC 25 data for all 3 markers, while others lacked information on ER, PR, or HER2 status (Table 1). The IHC data provided were used to calculate false-positive rates for each study, defined as tumors that were predicted negative for ER, PR, or HER2 by bimodal filtering of mRNA expression, but were positive by IHC. The overall false positive rates were 1.7%, 1.7%, and 0.9% for ER, PR, and HER2, respectively, 30 demonstrating that bimodal filtering of data sets by mRNA expression is a reliable method for identifying TNBC tumors from data sets lacking IHC information (Table WO 2013/075059 PCT/US2012/065724 - 33 1). The overall frequency of TNBC across the training data set was 16% and is consistent with the prevalence of TNBC previously reported in 2 other large studies performed on 3,744 cases (17%) (17) and 1,726 cases (16%) (18). k-means and consensus clustering reveal 6 TNBC subtypes. To identify 5 global differences in GE between TNBC subtypes, k-means clustering was performed on the most differentially expressed genes (SD>0.8). Using the silhouette width (s[i]) as a measure of relative closeness of individual samples to their clusters, k-means clustering classified 337 of the 386 TNBC tumors into 6 stable clusters (s[i] > 0) and 49 tumors into 1 unstable cluster (s[i] > 0) (Figure 2A). Clustering resulted 10 in a distribution of samples in all 7 clusters independent of each data set, n = 14, indicating that confounding factors such as batch effect, RNA amplification, and sample quality did not influence cluster distribution (Figure 8). Sample classification robustness was analyzed by consensus clustering, which involves k-means clustering by resampling (1,000 iterations) randomly selected tumor profiles. The consensus 15 matrix is a visual representation of the proportion of times in which 2 samples are clustered together across the resampling iterations (Figure 2B). Groups of samples that frequently cluster with one another are pictorially represented by darker shades of red. To determine the number of clusters present in the data, the area under the curve of the consensus distribution function (CDF) plot was examined (Figure 2C). 20 The point at which the area under the curve ceases to show marked increases with additional cluster number (k) indicates the ideal number of clusters (Figure 2D; Monti S. Machine Learning. 2003;52(1-2):91-118). Therefore, the optimal number of clusters is 7, as defined by the consensus plots consistent with the k-means clustering (6 stable, 1 unstable). Unsupervised dimension reduction by principal 25 component analysis demonstrated fundamental differences in GE between tumor subtypes identified by k-means and consensus clustering (Figure 2E). To assess whether similar TNBC subtypes could be generated from an independent TNBC cohort, 201 TNBC samples were compiled from bimodal filtering of 7 additional publicly available data sets (Table 2). A GE signature was 30 derived from the most differentially expressed genes found in the TNBC training set (see Methods) and used to predict which TNBC subtype was best-fit for each of the WO 2013/075059 PCT/US2012/065724 - 34 tumors in the validation set. Each sample from the validation set was assigned to 1 of the TNBC subtypes derived from the training data set based on the highest Pearson correlation and lowest P value. Samples with correlations differing by P < 0.05 were considered unclassifiable. The validation set resulted in proportioned 5 subtypes similar to those of the initial k-means clustering of the training set. After the analysis was completed, an ad hoc analysis was performed by combining the training and validation data sets (587 tumors), which resulted in 7 subtypes identified by consensus clustering (Figure 9), with similar enrichment in gene ontologies. This further validates the stability of the subtypes and shows that 10 increasing sample size does not change optimal cluster number. Evaluation of GE profiles of RNA obtained from laser-capture microdissection of tumor cells from 2 data sets (GSE-22513; GSE-28821; GES-28796, Table 1; GSE-5847, Table 2) showed a similar pattern of distribution across all 7 subtypes, suggesting that these subtypes are indeed representative primarily of GE resulting from epithelial tumor 15 cells rather than stromal components surrounding the tumor (i.e., inflammatory cells, myofibroblasts, etc.) (Figure 10). Distinct gene ontologies are associated with each TNBC subtype. As an independent method to analyze TNBC subtypes, Gene Set Enrichment Analysis (GSE-A) (Subramanian A, et al. Proc Natl Acad Sci U S A. 2005;102(43):15545 20 15550) was performed on all genes from both the training and validation sets to determine the top canonical pathways associated with each TNBC subtype. The top enriched genes (P < 0.05), identified from the training set used to predict the validation set, are displayed in a heat map in Figure 3. The top canonical pathways substantially overlapped between the training and validation sets for each subtype, 25 indicating that the subtypes were reproducibly enriched in unique pathways (Table 4). Our 7 TNBC subtypes were characterized on the basis of gene ontologies and differential GE and subsequently labeled as follows: basal-like 1 (BL 1); basal-like 2 (BL2); immunomodulatory (IM); mesenchymal (M); mesenchymal stem-like (MSL); luminal androgen receptor (LAR); and unstable (UNS) (Figure 3). 30 Independent analysis of 5 data sets based on TNBCs identified by IHC staining (n = WO 2013/075059 PCT/US2012/065724 - 35 183) resulted in 4 clusters with GE similar to that of basal-like, IM, mesenchymal like, and LAR subtypes (Figures 11 A, 11 B and 12). BLI and BL2 subtypes. The top gene ontologies for the BLI subtype are heavily enriched in cell cycle and cell division components and pathways (cell cycle, 5 DNA replication reactome, G2 cell-cycle pathway, RNA polymerase, and G 1 to S cell cycle) (Figure 3). The annotations are supported by the expression of genes associated with proliferation, such as A URKA, A URKB, CENPA, CENPF, BUB], TTK, CCNA2, PRC], MYC, NRAS, PLKJ, and BIRC5 (Figure 13). Elevated DNA damage response (ATR/BRCA) pathways accompany the proliferation pathways in 10 the BL1 subtype (Figure 3). Increased proliferation and cell-cycle checkpoint loss are consistent with the elevated expression of the DNA damage response genes observed (CHEK1, FANCA, FANCG, RAD54BP, RAD51, NBN, EXO1, MSH2, MCM10, RAD21, and MDC1) (Figure 13). The highly proliferative nature of this subtype is further supported by the 15 finding of high Ki-67 mRNA expression (MKI67) (Figure 13) and nuclear Ki-67 staining as assessed by IHC analysis (BL1 + BL2= 70% vs. other subtypes = 42%; P < 0.05) (Figures 14A-14B). Enrichment of proliferation genes and increased Ki-67 expression in basal-like TNBC tumors suggest that this subtype would preferentially respond to antimitotic agents such as taxanes (paclitaxel or docetaxel) (Chakravarthy 20 AB, et at. Clin Cancer Res. 2006;12(5):1570-1576; Bauer JA, et al. Clin Cancer Res. 2010; 16(2):681-690). This is indeed the case when comparing the percentage of patients achieving a pathologic complete response (pCR) in 42 TNBC patients treated with neoadjuvant taxane in 2 studies (Bauer JA, et al. Clin Cancer Res. 2010; 16(2):681-690; Juul N, et al. Lancet Oncol. 2010;11(4):358-365). In these 25 combined studies, TNBC patients whose tumors correlated to the basal-like (BLI and BL2) subtype had a significantly higher pCR (63%; P = 0.042) when treated with taxane-based therapies as compared with mesenchymal-like (31%) or LAR (14%) subtypes (Figure 15). The BL2 subtype displays unique gene ontologies involving growth factor 30 signaling (EGF pathway, NGF pathway, MET pathway, Wnt/p-catenin, and IGF1R pathway) as well as glycolysis and gluconeogenesis (Figure 3). Likewise, the BL2 WO 2013/075059 PCT/US2012/065724 - 36 subtype is uniquely enriched in growth factor receptor GE such as EGFR, MET, and EPHA2 (Figure 13). This subtype has features suggestive of basal/myoepithelial origin as demonstrated by higher expression levels of TP63 and MME (CD 10) (Figure 13). 5 IM subtype. The IM subtype is enriched for gene ontologies in immune cell processes. These processes include immune cell signaling (THI/TH2 pathway, NK cell pathway, B cell receptor [BCR] signaling pathway, DC pathway, and T cell receptor signaling), cytokine signaling (cytokine pathway, IL-12 pathway, and IL-7 pathway), antigen processing and presentation, and signaling through core immune 10 signal transduction pathways (NFKB, TNF, and JAK/STAT signaling) (Figure 3). The IM signaling is evidenced by immune signal transduction GE (Figure 13), in addition to immune cell-surface antigens, cytokine signaling, complement cascade, chemokine receptors and ligands, and antigen presentation (Figure 16). Since a similar proportion of samples that were microdissected fell into the IM subtype, it is 15 likely that the IM characteristics are unique to the tumor cells themselves and not a reflection of immune cell infiltrate (Figure 10). Immune signaling genes within the IM subtype substantially overlap with a gene signature for medullary breast cancer, a rare, histologically distinct form of TNBC that despite its high-grade histology is associated with a favorable prognosis (Bertucci F, et al. Cancer Res. 20 2006;66(9):4636-4644). M and MSL subtypes. The M subtype displays a variety of unique gene ontologies that are heavily enriched in components and pathways involved in cell motility (regulation of actin by Rho), ECM receptor interaction, and cell differentiation pathways (Wnt pathway, anaplastic lymphoma kinase [ALK] 25 pathway, and TGF-p signaling) (Figure 3). The MSL subtype shares enrichment of genes for similar biological processes with the M subtype, including cell motility (Rho pathway), cellular differentiation, and growth pathways (ALK pathway, TGF-P signaling and Wnt/-catenin pathway). However, unique to the MSL are genes representing components and processes linked to growth factor signaling pathways 30 that include inositol phosphate metabolism, EGFR, PDGF, calcium signaling, G .1 WO 2013/075059 PCT/US2012/065724 - 37 protein coupled receptor, and ERKl/2 signaling as well as ABC transporter and adipocytokine signaling (Figure 3). The prevalence of cell differentiation and growth factor signaling pathways is illustrated by expression of TGF-p signaling pathway components (TGFB1L1, 5 BGNSMAD6, SMAD7, NOTCH], TGFB1, TGFB2, TGFB3, TGFBRJ, TGFBR2, and TGFBR3), epithelial-mesenchymal transition-associated (EMT-associated) genes (MMP2, ACTA2, SNAI2, SPARC, TAGLN, TCF4, TWIST], ZEB], COL3A], COL5A2, GNG11, ZEB2, and decreased E-cadherin [CDH1] expression), growth factors (FGF, IGF, and PDGF pathways), and Wnt/p-catenin signaling (CTNNB1, 10 DKK2, DKK3, SFRP4, TCF4, TCF7L2, FZD4, CA V1, CA V2, and CCND2) (Figure 13). The MSL subtype is also uniquely enriched in genes involved in angiogenesis, including VEGFR2 (KDR), TEK, TIE1, and EPAS] as well as immune signaling evidenced by an overlap in GE unique to the IM subtype (Figure 16). One interesting difference between the M and MSL subtypes is that the MSL 15 subtype expresses low levels of proliferation genes (Figure 13). This decreased proliferation is accompanied by enrichment in the expression of genes associated with stem cells (ABCA8, PROCR, ENG, ALDHA1, PER], ABCB1, TERF2IP BCL2, BMP2, and THY]), numerous HOX genes (HOXA5, HOXA 10, MEIS], MEIS2, MEOX, MEOX2, and MSX1), and mesenchymal stem cell-specific markers (BMP2, 20 ENG, ITGA V, KDR, NGFR, NT5E, PDGFRB, THY], and VCAMI) (Figure 13). The signaling pathway components differentially expressed in the M and MSL groups share similar features with a highly dedifferentiated type of breast cancer called metaplastic breast cancer, which is characterized by mesenchymal/sarcomatoid or squamous features and is chemoresistant (Gibson GR, et al. Am Surg. 25 2005;71(9):725-730). The MSL subtype also displays low expression of claudins 3, 4, and 7, consistent with a recently identified claudin-low subtype of breast cancer (Figures 13 and 17; Prat et al., Breast Cancer Res, 2010; 12(5):R68). Hierarchical clustering of TNBC GE profiles using the claudin-low gene predictor set (n = 770) segregated 30 a portion of the M and MSL subtypes with low claudin (3, 4, and 7), cytokeratin (KRT7, KRT8, KRT18, and KRT19), and CD24 expression (Figure 17; Prat A, et al.

WO 2013/075059 PCT/US2012/065724 - 38 Breast Cancer Res. 2010;12(5):R68). This population of claudin-low-expressing tumors also had high expression of genes associated with EMT (FBNI, MMP2, PDGFRB, THY], SPARC, TGFBR2, PDGFRA, TWIST, CA V1, CAV2, and SERPINE]). 5 LAR subtype. GE in the LAR group was the most differential among TNBC subtypes. This subtype is ER negative, but gene ontologies are heavily enriched in hormonally regulated pathways including steroid synthesis, porphyrin metabolism, and androgen/estrogen metabolism (Figure 3). Whether other hormone-regulated pathways such as androgen receptor (AR) signaling, previously reported in ER 10 negative breast cancer (Hayes MJ et al. Clin Cancer Res. 2008;14(13):4038-4044), could be responsible for the GE patterns in this LAR subtype was investigated. Indeed, it was found that AR mRNA was highly expressed, on average at 9-fold greater than all other subtypes (Figure 13). Tumors within the LAR group also expressed numerous downstream AR targets and coactivators (DHCR24, ALCAM 15 FASN, FKBP5, APOD, PIP, SPDEF, and CLDN8) (Figure 13). In addition to AR mRNA expression, AR protein expression was investigated by IHC across all TNBC tumors from the Vanderbilt cohort (n = 20). The percentage of tumor cells scored with nuclear AR staining and the intensity of staining were significantly higher in the LAR subtype (>10-fold; P < 0.004) compared with all other TNBC subtypes 20 (Figures 18A- 18B). Hierarchical clustering was performed using an AR-activated gene signature (Chen CD, et al. Nat Med. 2004; 10(1):33-39) on the training and validation TNBC tumor data sets. Hierarchical clustering with this signature segregated the majority of LAR tumor profiles from other subtypes (Figures 19A 19B). Tumors in the LAR subtype display luminal GE patterns, with FOXAI, 25 KRT18, and XBP1 among the most highly expressed genes (Figure 20A). Others have previously described a breast cancer subgroup expressing AR termed molecular apocrine (Chen CD, et al. Nat Med. 2004; 10(l):33-39). While we do not have detailed pathology reports for all of the LAR tumors in this study, the TNBC subgroup GE centroids were used to correlate the apocrine samples from GSE-1561 30 (Farmer P, et al. Oncogene. 2005;24(29):4660-4671; Table 1). The GE profiles of all 6 apocrine tumors (GSM26883, GSM26878, GSM26886, GSM26887, WO 2013/075059 PCT/US2012/065724 - 39 GSM26903, and GSM269 10) described in the study strongly correlate with LAR, suggesting that the LAR TNBC subtype is composed of AR-driven tumors that include the molecular apocrine subtype. Intrinsic molecular breast cancer subtype classification of TNBC. Breast 5 cancers can be classified as luminal or basal-like, dependent on their expression of different cytokeratins. TNBC tumor subtypes display differential expression of both basal-like cytokeratins (KRT5, KRT6A, KRT6B, KRT14, KRT16, KRT1 7, KRT23, and KRT81) and luminal cytokeratins (KRT7, KRT8, KRT18, and KRT19) across the subtypes (Figures 20A-20B). The UNS, BL1, BL2, and M subtypes expressed 10 higher levels of basal cytokeratin expression, while tumors within the LAR subtype lacked basal cytokeratin expression and expressed high levels of luminal cytokeratins and other luminal markers (FOXA1 and XBP1) (Figure 20A). In addition to cytokeratin expression, breast cancers can be classified by an "intrinsic/UNC" 306-gene set into 5 major subtypes (basal-like, HER2-like, normal 15 breast-like, luminal A, and luminal B) (Hu Z, et al. BMC Genomics. 2006;7:96). Since TNBCs are largely considered basal-like, each of the 386 TNBC tumor profiles were correlated to the intrinsic gene set centroids of the 5 molecular subtypes, as previously described (Hu Z, et al. BMC Genomics. 2006;7:96). Tumors were assigned to 1 of the molecular subtypes based on the highest correlation 20 coefficient between each TNBC expression profile and the 5 molecular subtype centroids. This analysis resulted in 49% (n = 188) of the TNBC training set classified as basal-like, 14% (n = 54) as luminal A, 11% (n = 42) as normal breast like, 8% (n = 31) as luminal B, 5% (n = 19) as HER2, and 13% (n = 52) as unclassifiable (Figure 20B). This confirms that most TNBCs classify to the basal 25 like molecular subtype. Both the unstable tumors and BLI tumors correlated strongly to the basal-like intrinsic molecular classification (76% and 85%, respectively). However, the BL2, IM, and M subtypes only moderately correlated to the basal-like molecular class (31%, 58%, and 47%, respectively), with a portion of tumors unclassified (22%, 17%, and 18%, respectively) (Figure 20B). The M and 30 MSL subtypes displayed the largest portion of tumors classified as normal breast like (25%, and 46%, respectively). The BL2, M, and MSL subtypes were a mixture WO 2013/075059 PCT/US2012/065724 - 40 of classifications, suggesting the intrinsic classification system may not be suitable for characterizing these TNBC subtypes. The majority of TNBC tumors within the LAR subtype were classified as either luminal A or luminal B (82%), and none were classified as basal-like, further supporting the luminal origin of the LAR subtype 5 (Figures 20A-20B). While only 49% of the tumors were classified as basal-like according to the intrinsic gene set, IHC staining performed on the Vanderbilt subset of tumors (n = 25) showed that the majority (88%) of TNBCs stained positive for the basal cytokeratins CK5/6. Additionally, 56% of the Vanderbilt tumors stained positive for EGFR, similar to a previous study that found 56% of n = 929 pooled 10 from 34 studies that were positive for EGFR or CK5/6 (Yang XR, et al, J Natl Cancer Inst. 2011;103(3):250-263). There were no statistical differences between CK5/6 and EGFR staining across TNBC subtypes. Thus, the majority of TNBCs display a basal-like phenotype by IHC, while only approximately half correlate to the basal-like intrinsic gene set. 15 Patient relapse-free survival and distant-metastasis-free survival differs among TNBC subtypes. Relapse-free survival (RFS) between TNBC subtypes was significantly different (log-rank test; P = 0.0083) (Figure 21A), despite variability of therapy and duration of treatment. RFS was significantly decreased in the LAR subtype compared with the BLI (hazard ratio [Hr]= 2.9), IM (Hr = 3.2), and MSL 20 (Hr = 10.5) subtypes (P < 0.05) (Figure 21B). RFS was significantly decreased in the M subtype compared with BLI (Hr = 2.6) and IM (Hr = 2.9), while the MSL subtype had higher RFS than the M subtype (Figure 2 1B). Distant- metastasis-free survival (DMFS) did not vary between TNBC subtypes (log-rank test; P = 0.2176), but the M subtype had a significantly higher Hr compared with the BLI (Hr = 2.4, P 25 < 0.05) and IM (Hr = 1.9, P <0.06) subtypes (Figures 21C and 21D). The increased RFS in the absence of increased DMFS for patients in the LAR subtype suggests that recurrence is due to local relapse. Tumor size and grade were not significantly different among TNBC subtypes, but age at diagnosis was greater in the LAR subtype (P < 0.0001), potentially contributing to decreased RFS of this subtype 30 compared with other subtypes (Figures 22A-22B).

WO 2013/075059 PCT/US2012/065724 - 41 TNBC cell line models for TNBC subtypes. Cell line models would facilitate preclinical experiments to define differential drug sensitivity of the distinct subtypes within this heterogeneous disease. Using the bimodal filtering approach on ER, PR, and HER2 GE levels from 2 independent breast cancer cell line data sets (GSE 5 10890 and ETABM-157), 24 and 25 triple-negative cell lines were identified in the GSE-10890 and ETABM-157 GE data sets, respectively. Of the cell lines present in both data sets, nearly all had similar predictions of triple-negative status by bimodal filtering. Discrepancies in some cell lines (e.g., HCC1500 and HCC1937) may be the result of differences in culturing methods and/or loss of hormone receptor 10 expression over time in culture. Analysis of these 2 data sets identified 30 nonoverlapping TNBC cell lines. GE profiles from the cell lines were correlated to the centroids for each of the 6 TNBC subtypes. The majority of cell lines (27 of 30) were assigned to TNBC subtypes (Table 3), except for BT20, HCC1395, and SW527, which had low 15 correlations (<0.1) or were similar between multiple subtypes (P > 0.05). Of the 7 cell lines that contained known BRCA1 and BRCA2 mutations, 5 correlated with the BLI or BL2 subtypes (HCC1937, HCC1599, HCC2157, HCC3153, and SUM149PT), consistent with this subtype containing tumors with defects in DNA repair pathways (Table 3) and gene ontologies enriched for GE involved in cell 20 cycle and DNA repair functions (Figure 13). Additionally, cell lines in the BLI and BL2 subtypes were more genomically unstable, displaying significantly more chromosomal rearrangements than those in the mesenchymal subtypes (M and MSL) (average translocations, inversions, deletions = 34 versus 14, respectively; P < 0.01), as determined by SKY-FISH (www.path.cam.ac.uk/-pawefish/index.html) (Figures 25 23A-23B). All cell lines that correlated to the basal-like intrinsic molecular subtype were in the BLI, BL2, or IM subtypes (Table 3). Only 2 cell lines (HCC 1187 and DU4475) were placed into the IM subtype, suggesting that the IM subtype is underrepresented in cell culture. Cell lines in the M and MSL subtypes were generated from highly 30 dedifferentiated tumors derived from unique pathologies (e.g., HS578T, carcinosarcoma; and SUM159PT, anaplastic carcinoma) and expressed both WO 2013/075059 PCT/US2012/065724 - 42 epithelial and mesenchymal components. All cell lines assigned to the M and MSL subtype have spindle-like morphology in 2D culture (CAL-120, CAL-51, MDA MB-157, MDA-MB-231, MDA-MB-436, SUM159PT, HS578T, and BT549) or stellate-like morphology in 3D (Kenny PA et al., Mol Oncol, 2007; 1:84-96). Five 5 cell lines matched to the LAR subtype (MDA-MB-453, SUM185PE, HCC2185, CAL-148, and MFM-223). Two distinct basal groups (A and B) have been identified by GE profiling of breast cancer cell lines (Neve RM, et al. Cancer Cell. 2006;10(6):515-527). Basal A cell lines display epithelial characteristics and are associated with BRCA1 gene 10 signatures, while basal B cell lines are more invasive and display mesenchymal and stem/progenitor-like characteristics. The GE analyses revealed that the majority of basal A cell lines belong to the BLI and BL2 subtypes, while the majority of basal B cell lines fall into the M and MSL subtypes (Table 3). Hierarchical clustering analysis was tested on all TNBC cell lines using the most differentially expressed 15 genes from the tumors to determine whether GE patterns of the cell lines are similar within TNBC subtypes (Figure 24). Three clusters were identified: LAR, containing all 4 LAR lines; basal-like, containing lines in the BLI and BL2 subtypes; and mesenchymal-like, containing lines in the M and MSL subtypes. This clustering analysis indicates that TNBC cell lines can be classified into 3 main groups: basal 20 like (BL1 and BL2); mesenchymal-like (M and MSL); and LAR. This classification will be used in subsequent sections. TNBC cell lines have differential sensitivity to therapeutic agents. There are a variety of targeted therapies undergoing clinical investigation in patients with TNBC including those targeting PARP (Farmer H, et al. Nature. 25 2005;434(7035):917-921), AR (Agoff NS et al. Am J Clin Pathol. 2003;120(5):725-73 1), Src (Finn RS, et al. Breast Cancer Res Treat. 2007;105(3):319-326), and PI3K/mTOR (Marty B, et al. Breast Cancer Res. 2008;10(6):R1l01) signaling. The panel of TNBC cell lines was used to assess differential response to several agents targeting these pathways. For comparison 30 primary human mammary epithelial cells (HMECs) were analyzed in cell viability assays. The half-maximal inhibitory concentration (IC 50 ) values were determined for WO 2013/075059 PCT/US2012/065724 -43 the following drugs [targets]: veliparib (ABT-888) [PARP1/2]; olaparib (AZD2281) [PARPl/2]; cisplatin [DNA]; bicalutamide [AR]; 17-DMAG [Hsp9o]; dasatinib [Src/ Abl]; and NVP-BEZ235 [P13K and mTOR]. Cell-cycle and DNA damage response genes are elevated in the BLI and 5 BL2 subtypes (Figure 13), and 5 of the 7 lines in the cell line panel that carry BRCA1/2-mutations reside in these subtypes (Table 3). Given the previous clinical trial observations with PARP inhibitors and cisplatin (Bhattacharyya A et al. J Biol Chem. 2000;275(31):23899-23903; Jones P, et al. J Med Chem. 2009;52(22):7170 7185; Evers B, et al. Clin Cancer Res. 2008;14(12):3916-3925), it was predicted 10 that these agents would preferentially decrease viability in cell lines with DNA repair defects that are representative of the BLI and BL2 subtypes. The TNBC cell line panel was treated with the PARP inhibitors veliparib and olaparib, but not all cell lines representative of the basal-like TNBC subtypes were sensitive to PARP inhibition (Figures 4A-4D). The BRCAl-null cell line HCC 1937 was sensitive to 15 veliparib (ICso = 4 pM) but not olaparib (ICso > 100 piM), while the BRCA 1-mutant MDA-MB-436 was sensitive to both PARP inhibitors (veliparib IC 50 = 18 pIM and olaparib ICSO =14 pM). The BRCA2-mutant cell line HCC1599 lacked sensitivity to either PARP inhibitor (veliparib IC 50 > 30 [M and olaparib IC 50 > 100 pM). Thus, in addition to BRCA1/2 status, other properties of the tumor may dictate sensitivity to 20 a given PARP inhibitor. Unlike PARP inhibitor sensitivity, basal-like lines were significantly more sensitive to cisplatin than mesenchymal-like lines (average IC 50 = 8 [M vs. IC 5 o = 16 iM, P = 0.032) or LAR lines (average IC 50 = 8 pM vs. ICso = 15 ,M, P = 0,017) (Figure 4, E and F). The BRCAl-mutant cells (SUM149PT, HCC1937, and MDA-MB-436) and BRCA2-mutant cells (HCC1599) were among 25 the most sensitive to cisplatin treatment (Figure 4E). Since cell lines representative of the LAR subtype (MDA-MB-453, SUM185PE, CAL-148, and MFM-223) express high levels of AR mRNA and protein (Figure 25A), we compared the sensitivity of TNBC cell lines to the AR antagonist bicalutamide (Figures 5A and 5B). IC 50 values for the majority of TNBC 30 cell lines were not achieved using the highest dose of 500 M. However, all LAR cell lines tested (SUM185PE, CAL-148, MDA-MB-453, and MFM-223) and a WO 2013/075059 PCT/US2012/065724 - 44 subset of mesenchymal-like cell lines that express low levels of AR (HS578T, BT549, CAL-5 1, and MDA-MB-23 1) were more sensitive to bicalutamide than basal-like cell lines (average IC 50 = 227 ptM vs. IC 5 0 > 600 pM, P = 0.007; and average IC 50 =361 tM vs. IC 50 > 600 ptM, P = 0.038, respectively) (Figure 5A). 5 Since AR requires the Hsp90 chaperone for proper protein folding and stability (Solit DB, et al. Clin Cancer Res. 2002; 8(5):986-993), the sensitivity of the cell line panel to an Hsp90 inhibitor, 17-dimethylaminoethylamino-17 demethoxy-geldanamycin (17-DMAG), was determined. Again, the LAR cell lines were more sensitive to 17-DMAG compared with the majority of basal-like (average 10 IC 5 0 = 16 nM vs. IC 50 = 81 nM; P = 0.004) and mesenchymal-like (average IC 50 = 16 nM vs. IC 50 = 117 nM; P = 0.05) cell lines (Figures 5C and 5D), albeit 17 DMAG has many other targets. These data strongly suggest that LAR tumors are driven by AR signaling and AR represents a therapeutic target for this subtype. Importantly, AR status in TNBC patients represents a molecular marker for 15 preselection of patients for antiandrogen therapy. To determine whether the growth of the LAR cell lines was AR dependent, MFM-223, MDA-MB-453, and SUM1 85PE cell lines were treated with control and AR-targeting siRNA. Knockdown of AR in the experimental samples at the protein level was verified (Figure 25B), colony growth assays were performed, and the 20 number of colonies formed after 14 days of siRNA transfection were analyzed. The ability of all the LAR cell lines to form colonies was significantly reduced after knockdown of AR expression as compared with control samples (MFM-223 = 55%, P = 0.031; MDA-MB-453= 51%, P = 0.004; and SUM185PE= 42.3%, P = 0.002) (Figure 25C), indicating that AR expression is in part responsible for tumor cell 25 viability/survival. GE analysis of the mesenchymal-like subtypes demonstrated enrichment in the expression of genes that make up components and pathways associated with EMT (TGFP, ECM-receptor interaction, ALK, Wnt/p-catenin, and Rac1) and those associated with cell motility (focal adhesion, integrin signaling, RacI, striated 30 muscle contraction, and regulation of actin by Rho GTPase) Figure 3). Since the nonreceptor tyrosine kinase Src plays critical roles in cell migration and the WO 2013/075059 PCT/US2012/065724 - 45 mesenchymal-like subtypes are enriched in cell motility pathways, the effect of the Src inhibitor dasatinib on the panel of TNBC lines was analyzed. Cell lines belonging to the mesenchymal-like subtypes (M and MSL) were more sensitive to dasatinib than the LAR cell lines (average IC 50 = 22 tM vs. IC 50 =88 pM, P 5 0.024) (Figures 6A and 6B). Since activating mutations in PIK3CA are the most frequent genetic event in breast cancer (Samuels Y, et al. Science. 2004;304(5670):554), the TNBC cell lines were treated with the dual PI3K/mTOR inhibitor NVP-BEZ235 (Maira SM, et al. Mol Cancer Ther. 2008;7(7):1851-1863). TNBC cell lines that have activated 10 PI3K/AKT signaling due to PIK3CA mutations or PTEN deficiency (Table 3) were highly sensitive to NVPBEZ235 (Figure 6C). In addition, mesenchymal-like TNBC cell lines were more sensitive to NVP-BEZ235 compared with basallike cell lines (average IC 50 = 44 nM vs. IC 50 = 201 nM; P = 0.00 1) (Figures 6C and 6D), which may suggest that deregulation of the P13K pathway is important for this subtype. 15 LAR cell lines were also more sensitive to NVP-BEZ235 compared with basal-like cell lines (average ICso = 37 nM vs. 116 nM; P = 0.01) (Figures 6C and 6D). This sensitivity can be explained by PIK3CA mutations, frequent in the LAR subtype, with all LAR cell lines containing PIK3CA-activating mutations (HCC2185, MDA MB-453 CAL-148, MFM-223, and SUM185PE) (Table 3). While PIK3CA 20 mutations predicted NVP-BEZ235 sensitivity, PTEN deficiencies (mutation or loss of protein expression) did not correlate with sensitivity. Xenograft tumors derived from TNBC cell lines display differential sensitivity to therapeutic agents in vivo. In order to further analyze the susceptibility or resistance of TNBC subtypes to therapeutic agents in a more physiological setting 25 than 2D culture, xenograft tumors were established in nude mice from cell lines representative of the basallike (HCC 1806 and MDA-MB-468), mesenchymal-like (CAL-51 and SUM159PT), or LAR (SUM185PE and CAL-148) subtypes. After tumors reached an approximate volume of 25-50 mm 3 , the mice were treated with cisplatin, NVP-BEZ235 or bicalutamide. The sensitivity of the TNBC cell lines to 30 the therapeutic agents when grown as 3D xenograft tumors in vivo was very similar to that seen with the cell lines grown in 2D monolayer culture. The xenograft tumors WO 2013/075059 PCT/US2012/065724 - 46 derived from the 2 cell lines representative of basal-like tumors (HCC1806 and MDA-MB-468) were highly and differentially sensitive to cisplatin and were significantly growth inhibited (P < 0.0001) relative to treatment with vehicle control or the other experimental treatments (bicalutamide and NVP-BEZ235) (Figure 7). 5 The tumors derived from the LAR cell lines (SUM185PE and CAL-148) and the mesenchymal-like cell line that expresses low level AR protein (CAL-51) (Figures 25A-25C) displayed significant sensitivity to bicalutamide (Figure 7). The xenograft tumors derived from all the cell lines carrying activating PIK3CA mutations (SUM185PE, CAL-148, CAL-51, and SUM159PT) had partial response or complete 10 compared with NVP-BEZ235 (Figure 7). In summary, cell lines representative of the 6 TNBC subtypes display different sensitivities to a variety of agents, and importantly, these differences can be attributed to distinct expression of cellular components and presence of mutations in key oncogenes and tumor suppressors. The results have immediate clinical 15 translation, as they provide a valuable platform and insights for ongoing and future clinical investigation. The data inidcate that patients with basal-like TNBC should be treated with agents that engage DNA damage signaling response pathways; those with tumors expressing AR should receive bicalutamide alone or in combination with P13K inhibitors; and those with mesenchymallike TNBC should be considered 20 for trials exploring the activities of an Src antagonist in combination with a P13K inhibitor. Supplemental Results K-means Clustering of a subset of five datasets that were identified as TNBC 25 by IHC. Since TNBCs are not identified by GE in the clinical setting, a similar GE analysis was performed on a dataset from TNBCs identified using IHC (n=183) complied from five datasets: GSE7904 (n= 22), GSE19615 (n= 34), GSE20194 (n= 67), E-TABM-158 (n= 30) and GSE22513, GSE-XXX (n= 30). This compilation included 136 tumors (74.3%) that were identified by our bimodal filtering analysis. 30 K-means clustering performed on the most differentially expressed genes (SD> 0.8) resulted in the identification of 5 TNBC subtypes (Figures 11 A- I1B). One subtype WO 2013/075059 PCT/US2012/065724 - 47 contained only three tumors and analysis of GE showed these tumors to have high levels of ER and PR Figure 1 IB). These tumors were considered to be false negatives by IHC and were removed from further analysis. Comparison of these IHC-identified TNBC samples using genes differentially expressed from the six 5 TNBC subtypes revealed similar patterns of gene enrichment (Figure 12). Discussion Through GE analysis of 3247 breast cancers, demonstrate herein is that 10 TNBCs can be reliably identified by filtering GE profiles for ER, PR, and HER2 mRNA levels. 587 TNBC GE profiles from 21 studies (training set = 386 and validation set = 201) were compiled. An 18% incidence rate of TNBC across the 21 independent studies (training and validation combined) was observed, similar to previously reported TNBC prevalence (Bertucci F, et al. Int J Cancer. 15 2008;123(1):236-240; Rakha EA, et al. Clin Cancer Res. 2009;15(7):2302-2310). k-means and consensus clustering of tumor profiles revealed that TNBC is composed of 6 stable subtypes enriched in distinct gene ontologies and GE patterns. Furthermore, using a GE signature derived from TNBC patient tumors, cell-line models for each of the TNBC subtypes were identified. 20 Previously, the majority (50/o-90%) of TNBCs have been classified as basal-like either by IHC or by correlation to the intrinsic molecular breast cancer subtypes (Bertucci F, et al. Int J Cancer. 2008;123(1):236-240; Rakha EA, et al. Clin Cancer Res. 2009;15(7):2302-23 10; Kreike B, et al. Breast Cancer Res. 2007;9(5):R65). A previous TNBC study identified 5 distinct hierarchical clusters in which 91% (88 of 25 97) of TNBCs identified by IHC correlated to the basal-like subtype (Kreike B, et al. Breast Cancer Res. 2007;9(5):R65). However, the study lacked molecular analysis of the tumors and conclusions were limited to clinical outcomes based on pathological markers. The relationship between TNBC and basal-like breast cancer remains controversial (Rakha E et al. Clin Cancer Res. 2008;14(2):618). The 30 proportion of TNBCs with basal-like GE in our study was 47%, resulting in a higher proportion of TNBCs that correlate with other molecular subtypes: luminal A (17%), WO 2013/075059 PCT/US2012/065724 -48 normal breast-like (12%), luminal B (6%), HER2 (6%), or unclassified (12%). The study herein indicates that TNBC is not limited to tumors with a basal-like phenotype; rather it is a heterogeneous collection of tumors with distinct phenotypes, as evidenced by the diverse GE patterns and varying sensitivity of 5 representative cell lines to the targeted therapies assessed in this study. The BLI and BL2 subtypes express high levels of genes involved in cell proliferation and DNA damage response, suggesting patients with basal-like tumors would benefit from agents that preferentially target highly proliferative tumors (e.g., anti-mitotic and DNA-damaging agents). Consistent with this notion, patients with 10 basal-like tumors had up to a 4-fold higher pCR after taxane-based and radiation based treatment as compared with patients with tumors that displayed characteristics of the ML or LAR subtypes (Bauer JA, et al. Clin Cancer Res. 2010; 16(2):681-690; Juul N, et al. Lancet Oncol. 2010; 11(4):358-365). Nearly all of the cell lines with known mutations in BRCA1 and BRCA2 had 15 GE patterns that correlated with the basal-like subtype, which is in agreement with the current view that BRCA-mutant tumors display a basal-like phenotype (Stefansson OA, et al. Breast Cancer Res. 2009;1 1(4):R47). A number of non BRCAmutant cell lines that correlated with the basal-like TNBC subtypes contained nearly 2-fold the number of chromosome rearrangements as all other subtypes. 20 These findings suggest that a predominant characteristic of basal-like TNBC is genomic instability (Kwei KA et al.Mol Oncol. 2010;4(3):255-266). Because of GE similarities between basal-like TNBC and BRCA1-mutation carriers, PARP inhibitors are currently being tested in clinical trials for TNBC (Silver DP, et al. J Clin Oncol. 2010;28(7):1145-1153). Despite success in BRCA-null cells, there is 25 not sufficient evidence for PARP inhibitor efficacy in BRCA1/2-mutant breast cancer cells (Farmer H, et al. Nature. 2005;434(7035):917-921). PARP sensitivity across a panel of TNBC cell lines did not correlate with TNBC subtype alignment or BRCA1/2-status in this study using short-term cell viability assays (72 hours). However, the synthetic lethal effects of PARP inhibition in homologous 30 recombination-deficient cells may require multiple cell divisions and may be more appropriately tested in long-term survival assays. Other studies show that a subset of WO 2013/075059 PCT/US2012/065724 - 49 BRCA 1-mutant tumors lack large-scale genomic alterations and these tumors represent a distinct disease entity and may not be susceptible to PARP inhibition (Stefansson OA, et al. Breast Cancer Res. 2009;1 1(4):R47). Previous analyses of PARP inhibitors in isogenic BRCA2+/+ and BRCA2-/- cell lines suggest that 5 sensitivity to this targeted therapy is dependent on the molecular context of the DNA repair machinery and that DNA repair requiring homologous recombination involves multiple redundant pathways (Farmer H, et al. Nature. 2005;434(7035):917-921; Fong PC, et al. N Engl J Med. 2009;361(2):123-134). BRCA1 is a relatively large protein (1863 aa) that forms numerous complexes that may not be entirely disrupted 10 when BRCA1 and BRCA2 are mutated, as opposed to BRCA-null cells. Despite an incomplete understanding of the molecular mechanism of the PARP inhibitors in vivo, these drugs have been proven to be highly effective in the clinical trial setting (41% objective response rate; 6.2 month progression-free survival) (Audeh MW, et al. J Clin Oncol. 2009;27:5500). 15 It was also found that BRCAl -mutant and non-BRCA-mutant basal like cell lines had relatively higher sensitivity to cisplatin treatment compared with all other TNBC subtypes.The results are consistent with the observed 21% pCR in a clinical trial investigating neoadjuvant cisplatin as a single agent in a heterogeneous TNBC patient population (Garber JE, et al. Breast Cancer Res Treat. 2006;105:Sl49; Telli 20 ML, Ford JM. Clin Breast Cancer. 2010;10 supply 1:E16-E22). These data collectively suggest that the use of proliferation biomarkers such as Ki-67 and development of markers identifying defects in DNA damage response signaling could provide patient selection and tailored treatments for basal-like TNBC. Use of cisplatin as a "targeted" agent alone or in combination with antimitotics (taxanes) 25 and/or radiation may benefit patients with this subtype and a current trial is underway with these agents (Mayer IA, et al. J Clin Oncol. 2010;28(1):15). The IM subtype is highly enriched in immune cell signaling. Other studies have described the presence of immune response gene signatures in ER-negative and medullary breast cancers (Bertucci F, et al. Cancer Res. 2006;66(9):4636-4644; 30 Teschendorff AE et al. Genome Biol. 2007;8(8):R157). Similar to these studies, it was found elevated expression of T cell-associated genes, immune transcription WO 2013/075059 PCT/US2012/065724 - 50 factors, IFN regulatory factors, TNF, complement pathway, and antigen processing. It cannot be ruled out that the GE profile of the IM subtype comprises, at least in part, stromal components including immune cell infiltrate. However, the finding that the same proportion of microdissected tumors belongs to this group argues against 5 stromal contamination. The M and MSL subtypes share similar gene ontologies and GE profiles involving TGF-p, mTOR, Rac1/Rho, Wnt/p-catenin, FGFR, PDGFR, and VEGF signaling pathways. These signaling pathways are prominent in processes of EMT and stem cell-like properties of a CD44+CD24- population of normal mammary 10 cells (Mani SA, et al. Cell. 2008;133(4):704-715; Shipitsin M, et al. Cancer Cell. 2007; 11(3):259-273). Similarly, the MSL subtype is made up at least in part by the recently described claudin-low tumors, which lack luminal differentiation markers, high enrichment for EMT markers, immune response genes, and cancer stem cell like features (Prat A, et al. Breast Cancer Res. 2010;12(5):R68). Interestingly, the M 15 and MSL subtypes differed clinically, with patients in the M subtype presenting with shorter RFS. This may be a reflection of differences in proliferation, as the M subtype displayed higher expression of proliferation-associated genes, including Ki 67. Additionally, patients with the M and MSL subtypes had decreased 5-year DMFS, consistent with enrichment in pathways associated with metastasis and 20 motility. Tumors within the mesenchymal-like subtypes have GE profiles that are similar to those from mesenchymal cells and metaplastic breast cancers (Hennessy BT, et al. Cancer Res. 2009;69(10):4116-4124). Metaplastic breast cancers have lineage plasticity, including foci of spindle cell, as well as osseous or cartilaginous 25 differentiation (Gibson GR, et al. Am Surg. 2005;71(9):725-730). A recent study found that 47% of metaplastic breast cancers sequenced have PIK3CA mutations and have higher phosho-AKT expression (Hennessy BT, et al. Cancer Res. 2009;69(10):4116-4124). It was found herein that TNBC mesenchymallike cell lines preferentially responded to the dual PI3K/mTOR inhibitor NVP-BEZ235. This 30 response to NVP-BEZ235 was demonstrated in cell lines that carry PIK3CA mutations in xenografts in addition to an MSL cell line (SUM159PT) that lacks WO 2013/075059 PCT/US2012/065724 -51 PIK3CA mutation or PTEN deficiencies, suggesting that the PI3K/mTOR pathway is important in the mesenchymal-like subtype. The mesenchymal-like subtypes were enriched in pathways associated with EMT and cell motility. There is evidence of a prominent role for Src in tumor cells 5 that are highly invasive, such as those that have undergone EMT (Guarino M. J Cell Physiol. 2010;223(1):14-26). Accordingly, we found that mesenchymal-like TNBC cell lines had differential sensitivity to dasatinib. Markers of EMT may have clinical value for patient preselection for trials using dasatinib. In addition, Wnt-signaling pathways regulate EMT and may contribute to tumor cell invasion (Shin SY, et al. 10 Cancer Res.2010;70(17):6715-6724). Mutations in the Wnt/p-catenin pathway (CTNNB 1, APC, and WISP3) occur frequently (52%) in metaplastic breast cancer, suggesting that deregulated Wnt/p-catenin pathway in these tumors may be a viable therapeutic target (Hayes MJ et al. Clin Cancer Res. 2008;14(13):4038-4044). Inhibitors of Wnt/p-catenin are of great interest and currently are in preclinical 15 development (MacDonald BT, et al. Dev Cell. 2009;17(1):9-26). Drugs targeting this pathway could be of value for treating mesenchymal-like TNBC. The LAR subtype was readily subclassified by an AR gene signature and high levels of luminal cytokeratin expression. GE analysis of the LAR subtype is consistent with a prior report of a subset of ER-negative tumors expressing AR 20 regulated genes (Doane AS, et al. Oncogene. 2006;25(28):3994-4008). In addition, Farmer et al. described an apocrine tumor subtype based on GE profiling that was characterized by AR expression distinguishing this tumor subtype from other basal like tumors (Farmer P, et al. Oncogene. 2005;24(29):4660-467 1). In the GE analysis of tumors from 21 studies described herein, the prevalence of the LAR tumors was 25 11% (62 of 587) of TNBCs or 2% (62 of 3247) of all breast cancers. Analysis of clinical data demonstrated that patients in the LAR subtype had higher RFS but no difference in DMFS compared with all other TNBC subtypes, suggesting these patients have local relapse. The higher RFS could imply that this group of patients received ineffective therapies (standard chemotherapy); however, patients in the 30 LAR group were significantly older at diagnosis and the extent of disease or age associated comorbidities that affect the ability to deliver treatment as planned may WO 2013/075059 PCT/US2012/065724 - 52 have contributed to relapse. Older age at diagnosis has previously been reported in patients with AR-positive TNBC and is associated with postmenopausal status (Agoff NS et al. Am J Clin Pathol. 2003;120(5):725-73 1). Whether this AR-driven subtype is arising from hormone-replacement therapy (HRT) merits further 5 investigation; however, it is becoming clear that the risk of breast cancer increases with HRT, and synthetic progestins such as medroxyprogesterone acetate have been shown to bind and disrupt AR (Birrell SN et al.FASEB J. 2007;21(10):2285-2293). As described herein, five (5) cell lines were identified that represent the LAR subtype and it was shown that they are sensitive to bicalutamide and 17-DMAG, 10 suggesting that therapies targeting AR may be effective against tumors that express this hormone receptor. In fact, there is a clinical trial (NCT00468715) underway testing the effect of bicalutamide in preselected patients with ER/PR-negative AR positive tumors. This further supports using in silico-based approaches to provide leads for trials with other targeted therapies in TNBC subtypes. 15 In addition to agents that target AR function, LAR cell lines were also sensitive to P13K inhibition. This sensitivity correlated with PIK3CA mutations. All 5 LAR cells lines have activating PIK3CA mutations and are sensitive to the P13K inhibitor NVP-BEZ235, similar to ER-positive breast cancer in which PIK3CA mutations are common (Gonzalez-Angulo AM, et al. Clin Cancer Res. 20 2009;15(7):2472-2478; Stemke-Hale K, et al. Cancer Res. 2008; 68(15):6084 6091). These findings suggest simultaneous targeting of AR and the PI3K/mTOR pathway may be of clinical benefit for LAR TNBC patients, as this combination has been shown to be synergistic in AR-dependent prostate cancer cells (Liu X et al. J Clin Oncol. 2010;28:(suppl; abstr e15049)). 25 The GE analysis of TNBC described herein demonstrates that with sufficient sample size, distinct subtypes of TNBC can be identified with putative molecular targets. The analyses provides biomarkers that can be used for patient selection in the design of clinical trials for TNBC as well as identification of potential markers of response to treatment. The identification of cell lines representing TNBC tumor 30 subtypes provides key models for preclinical studies with newly developed targeted agents.

WO 2013/075059 PCT/US2012/065724 - 53 Example 2 Frequent PIKCA mutations in AR-positive TNBC confer sensitivity to P13K and AR inhibitors Described herein is the further investigation of the molecular features of the 5 LAR subtype to identify rationale combinations of drugs that would show robust efficacy against AR-positive TNBC cells with the added goal of generating pre clinical data for rationale clinical trial design. It was discovered that PIK3CA kinase mutations were a frequent event in AR-positive TNBC tumors. It was found that genetic or pharmacological targeting of AR in LAR cells increased the therapeutic 10 benefit of PI3K/mTOR inhibition. Further, it was discovered that after bicalutamide treatment of mice bearing LAR xeongraft tumors, the P13K pathway signaling increased. Thus, the combination of AR antagonism and PI3K/mTOR inhibition was examined and an additive or synergistic effect, depending on the TNBC cell line, was found. Given that AR expression is a robust biomarker for selection of LAR 15 TNBC patients, the preclinical findings provide the rationale for preselecting AR positive TNBC patients for treatment with AR inhibitors and/or Pl3K/mTOR inhibitors, and future clinical trials exploring the efficacy of combinations of drugs that target AR and P13K signaling. 20 Results and Discussion AR-positive TNBC tumors are enriched for PIK3CA kinase domain mutations. High frequency PIK3CA mutations in AR-positive TNBC cell lines are described in Example 1. To determine if PIK3CA mutation was the result of in vitro selection during establishment of the tumor-derived cell lines or if PIK3CA 25 mutations were frequent in AR-positive TNBC tumors, performed Sanger sequencing was performed on 26 AR-positive TNBC cases. PCR-amplified regions from exons 9 and 20 that harbor the most frequently occurring activating mutations in PIK3CA were sequenced. Consistent with the observations in cell lines, nearly all of the detected mutations occurred at amino acid H1047 (20 of 21) with only one 30 occurring at E545K. Mutation status was confirmed with digital droplet PCR on DNA amplified from exon 20. Using both Sanger sequencing of exons 9 and 20 and WO 2013/075059 PCT/US2012/065724 - 54 digital droplet PCR, PIK3CA mutations were significantly (P<0.0056) enriched in AR-positive TNBC tumors (17 of 26, 65.3%) versus AR-negative TNBC tumors (10 of 26, 38.4%). 5 Validation of PIK3CA mutations in AR-Expressing TNBC To verify that PIK3CA mutations were enriched in AR-expressing TNBC, RNA-seq, DNA-seq and reverse-phase protein array (RPPA) data available from The Cancer Genome Atlas (TCGA) breast cohort (Cancer Genome Atlas Network, Nature, 490:61-70 (2012)) was analyzed. Following removal of false negatives for 10 ER, 66 TNBC samples were identified and classified according to molecular subtypes (Chen et al., Cancer informatics, 11:147-156 (2012)). The distribution of samples across the TNBC molecular subtypes was similar to that previously published (Lehmann et al., J Clin Invest, 121:2750-2767 (2011)). (BLl= 14 (21.1%), BL2 = 6 (9.0%), IM= 15 (22.7%), M = 17 (25.7%), MSL=7 (10.6%) and 15 LAR =7 (10.6%) (Figure 26B). As expected, RNA-seq data confirmed significant levels of AR RNA (6.8 vs. 0.3, p<0.0001) and protein (1.7 vs. 0.35, p<0.0001) in the LAR subtype. Overall, PIK3CA mutations were relatively rare in TNBC cases in the TCGA dataset (4 of 66, 6.0%). However, when sorted by subtype, PIK3CA mutations were significantly enriched in LAR TNBC (3 of 7, 42.8%) compared to 20 all other subtypes (1 of 59, 1.7%, p<0.0001), supporting our analysis of 26 TNBC cases above (Figure 26B). LAR TNBC cell lines have increased P13K pathway activation and are sensitive to P13K and PI3K/mTOR inhibition. 25 RPPA was used to investigate AR, PTEN, active AKT (pS473 and PT308) and active GSK3p (pS21/pS9) protein levels in LAR cell lines to determine if the P13K pathway was active in AR expressing, P13K mutant TNBC. Activated P13K signaling correlated with PIK3CA mutation and PTEN loss. AR-expressing TNBC cell lines expressed higher levels of active AKT as evidenced by increased pS473 30 (2.18 vs. -0.23, p=0.007) and pS308 (1.45 vs. -0.10, p=0.0032) and increased downstream phosphorylation of GSK3P at pS9 (1.00 vs. -0.26, p=0.0094) and WO 2013/075059 PCT/US2012/065724 - 55 pS21/pS9 (1.12 vs. -0.19, p=0.

0 0 2 6 ) (Figure 27). Similarly, those cell lines that were null for PTEN (BT549, CAL51, HCC1395, HCC1937, HCC38, HCC70, MDA-MB 436 and MDA-MB-468) displayed higher levels of active P13K signaling (p-AKT S473 1.28 vs. -1.28, p=0.0001, p-AKT T308 0.41 vs. -0.53, p=0.0018, p-GSK S9 5 0.37 vs. -0.76, p=0.0002 and pGSKap S9/S21 0.26 vs. -0.59, p=0.0009). To validate the RPPA data, immunoblot analysis of the PI3K/mTOR pathway in four of the LAR cell lines, as well as primary cultures of normal mammary epithelial cells (HMEC) and the AR-dependent prostate cancer cell line LNCaP (Figure 27), was also performed. AR receptor protein in the LAR TNBC cell 10 lines was expressed at levels equal to or greater than LNCaP cells. AR-expressing cell lines displayed higher levels of phosphorylated AKT (S473) and phosphorylated ribosomal S6, confirming an activated P13K pathway and downstream mTOR activation, respectively (Figure 27). To evaluate the clonality of AR expression, immunohistochemistry (IHC) for 15 AR and p-AKT was performed on a cell line-based array (Figure 27). The AR expressing cell lines displayed differing percentage of cells positive for AR protein expression (CAL148: 20%, MDA-MB-453: 50%, MFM-223: 50% and SUM185: 90%). However, p-AKT expression was much more homogenous, indicating that activation of PI3K/AKT signaling was an earlier clonal event. 20 To determine the sensitivity of PIK3CA mutation-bearing LAR cell lines to P13K inhibitors relative to other TNBC cell lines, a large panel of TNBC cell lines was treated with pan-PI3K inhibitors BKM120 (Novartis) and GDC-0941 (Genentech) or with the dual PI3K/mTOR inhibitors NVP-BEZ235 (Novartis) and GDC-0980 (Genentech). AR-expressing and other TNBC cell lines containing 25 PIK3CA mutations were among the most sensitive to P13K inhibition, as indicated by low half maximal inhibitory concentration (IC50) values (Figures 28A-28C). In contrast, PTEN deficiencies, while conferring P13K pathway activation, did not predict for sensitivity to P13K inhibitors (Figures 28A-28C). Nearly all of the PIK3CA mutant cell lines were more sensitive to PI3K/mTOR inhibitors than 30 primary cultures of human mammary epithelial cells (HMEC).

WO 2013/075059 PCT/US2012/065724 -56 Genetic and pharmacologic inhibition of AR increases the efficacy of P13K inhibition. Whether the combined targeting of AR and P13K would be more effective than either agent alone was investigated. To determine the effectiveness of 5 combinatorial treatment both genetic knockdown and pharmacological inhibition of AR with shRNA and bicalutamide, respectively, in the presence and absence of P13K inhibitors were performed (Figures 29A-29B). Immunoblot analysis confirmed decreased AR protein levels in three AR-expressing TNBC cell lines infected with lentiviral particles containing non-targeting shRNA or two AR 10 targeting hairpins (Figure 29A). Both shRNA expression vectors targeting AR decreased cell viability across multiple doses of P13K or PI3K/mTOR inhibitors (Figure 29B). Efficacy of AR-targeting in combination with P13K or PI3K/mTOR inhibitors was independent of compound, as similar results were obtained with BKM-120 and NVP-BEZ235 (Figure 34). 15 Given that knockdown of AR increased the efficacy of P13K inhibitors, pharmacological targeting of the AR was performed using bicalutamide. Both GDC-0941 and GDC-0980 displayed efficacy in AR-expressing TNBC cell lines alone and viability could be decreased further with the addition of 25pM bicalutamide (CDX) across multiple concentrations of P13K inhibitors (Figure 30A). 20 P13K inhibition in combination with bicalutamide decreased cell viability in an additive or synergistic manner, as demonstrated by viabilities at, or below, the line of additivity. Immunoblot analysis was performed on cell lystates from untreated cells or cells treated alone or in combination with CDX, GDC-0941 and GDC-0980 to evaluate attenuation of the PI3K/mTOR pathway. Both GDC-0941 and GDC 25 0980 decreased activated AKT (p-S473), however, GDC-0980 was more effective at decreasing mTOR activity as measured by the decrease in levels of phosphorylated S6 (Figure 30B). In contrast, bicalutamide (CDX) had minimal effect on either the PI3K/mTOR pathway or AR levels. Interestingly, GDC-0980 alone decreased AR levels and this decrease could be increased further by the addition of bicalutamide, 30 suggesting that mTOR signaling contributes to the stability of AR protein, similar to WO 2013/075059 PCT/US2012/065724 - 57 crosstalk observed in prostate cancer (Figure 30B) (Carver et al., Cancer Cell, 19:575-586 (2011)). To determine if the decreased viability observed in AR-positive TNBC cell lines that were treated with the combination of pathway inhibitors described above 5 was attributed to apoptotic cell death, activated caspase 3/7 levels after drug treatments were measured (Figures 3 1A-3 1B). Cells were treated with a high dose of adriamycin (ADR, 3 [ M) to measure the ability of cells to engage the apoptotic pathway. With exception of MFM-223 cells, each of the cell lines underwent robust caspase activation after 48 h of ADR treatment. While bicalutamide treatment did 10 not induce caspase activation, treatment with both GDC-0941 and GDC-0980 alone caused caspase activation and the levels of apoptosis could be increased in combination with bicalutamide (Figure 3 1A). In addition to caspase activation, the percentage of sub 2N DNA (sub-G1) containing cells (indicative of late stage apoptotic DNA fragmentation) was 15 quantitated by FACs analysis. While bicalutamide treatment did not result in caspase activation, there was an increase in the sub-GI fraction 48 h after treatment (5.25% to 18.44%) (Figure 31B). Treatment with GDC-0941 or GDC-0980 resulted in higher sub-Gi fraction that increased in combination with bicalutamide, consistent with the caspase activation described above (Figure 3 1B). Therefore, the decreased 20 viability of AR-positive TNBC cell lines treated with a combination of AR and P13K inhibitors can be, in part, attributed to apoptosis. Simultaneous targeting of AR and P13K decreases viability of AR-expressing cell lines grown in a 3-D forced suspension assay. 25 Since cell viability and drug sensitivity can be influenced by 2-D versus 3-D growth conditions, combined targeting of AR and PI3K/mTOR was evaluated in a 3-D forced suspension assay devoid of matrix. Cell lines were grown in forced suspension in agarose coated 96-well plates and treated with GDC-0941 or GDC0980 alone or in combination with bicalutamide (Figures 32A-32C). 30 Increasing doses of GDC-0941 and GDC-0980 alone decreased cell aggregate size and viability in AR-expressing TNBC cell lines to a lesser degree when compared to WO 2013/075059 PCT/US2012/065724 - 58 similar treatments in 2-D (Figures 32A and 32B). However, the addition of bicalutamide was either additive or synergistic with PI3K/mTOR inhibition (Figure 32B). Similar results were obtained when P13K was inhibited with BKM 120 or P13K and mTOR were targeted with NVP-BEZ235 (Figure 35). Together, these 5 results indicate that simultaneous targeting of AR and PI3K/mTOR is more effective than either agent alone in AR-expressing TNBC. P13K is activated after pharmacological inhibition of AR. As described in Example 1, evaluation of bicalutamide in AR-expressing 10 TNBC xenografts demonstrated significant reduction in tumor burden, however over time tumor growth occurred in the presence of drug, consistent with evolution of bicalutamide-resistant tumor cells. Since crosstalk between AR and P13K has been demonstrated in prostate cancer, whether the level of P13K activity changed between untreated and bicalutamide-resistant xenograft tumor cells was examined. Phospho 15 AKT (S473) levels were elevated in bicalutamide-resistant xenograft tumor cells (Figure 37). Consistent with these findings were the findings herein of dose dependent increases in p-AKT in cell lines upon genetic silencing of AR by shRNA (Figure 38B). 20 Simultaneous inhibition of AR and P13K signaling decreases the growth of LAR cell line-derived xenograft tumors. In order to further validate and investigate the combined targeting of AR and P13K in AR-expressing TNBC in vivo, xenograft tumors in nude mice was established using two representative AR-positive TNBC cell lines (CAL-148 and 25 MDA-MB-453). Following tumor establishment (25-50 mm 3 ), mice were either treated with vehicle, bicalutamide, NVP-BEZ235, GDC-0980 alone or either NVP BEZ235 or GDC-0980 in combination with bicalutamide. While each single agent alone significantly decreased tumor growth, both PI3K/mTOR inhibitors in combination with bicalutamide inhibited tumor growth to the greatest extent (Figure 30 32C). Thus, the combined inhibition of AR and P13K appears to be a rational WO 2013/075059 PCT/US2012/065724 -59 treatment strategy for AR-expressing TNBC and should be considered in future clinical trial design. Discussion 5 Herein we show that activating mutations in PIK3CA are a frequent event in a subset of AR-expressing TNBC (LAR) that display luminal gene expression patterns. The data herein provide the preclinical rationale for targeting AR in combination with PI3K/mTOR inhibition. Higher levels of p-AKT in cell line xenografts after prolonged treatment with 10 bicalutamide and increased p-AKT after siRNA knockdown of AR in cell lines is demonstrated herein. The data show that combined targeting of AR and P13K pathway inhibition can overcome the P13K pathway activation after AR inhibition. These studies demonstrate that both pathways cross-regulate each other by reciprocal feedback and coordinately support survival when either is repressed. The 15 feedback between the AR and P13K pathway may, in part, explain the enrichment of PIK3CA mutations observed in AR-positive TNBC, as this co-evolution may be necessary to overcome the increased PTEN levels driven by AR signaling. Methods 20 PIK3CA mutation evaluation. Two independent methods were used to identify PIK3CA mutations in tumor samples. Samples were identified as mutant if they had greater than 5% of the cells positive by microdroplet PCR, as there was complete concordance with Sanger sequencing calls, or called mutant by sanger when microdroplet samples were <5% 25 and >1% (Figures 33A-33B). Sanger To detect mutations in PIK3CA (H1047), a 588 bp fragment was PCR amplified from exon 20 of PIK3CA using the following amplification primers; 30 foreword: TGACATTTGAGCAAAGACCTG (SEQ ID NO: 1), reverse: CATAACATGAAATTGCGCATT (SEQ ID NO: 2). Sanger sequencing was WO 2013/075059 PCT/US2012/065724 - 60 performed on the PCR amplicon using the following internal sequencing primers; Exon Forward: ACATCATTTGCTCCAAACTGA (SEQ ID NO: 3) Reverse: CCTATGCAATCGGTCTTTGC (SEQ ID NO: 4) (Samuels et al., Science, 304:554 (2004)). 5 Droplet Digital PCR The TaqMan PCR reaction mixture was assembled from a 2x ddPCR Mastermix (Bio-Rad), 20x primer, and probes (final concentrations of 900 and 250 nM, respectively) and template (variable volume) in a final volume of 20 pL. Each 10 assembled ddPCR reaction mixture was then loaded into the sample well of an eight channel disposable droplet generator cartridge (Bio-Rad). A volume of 70 pL of droplet generation oil (Bio-Rad) was loaded into the oil well for each channel. The cartridge was placed into the droplet generator (Bio-Rad). The cartridge was removed from the droplet generator, where the droplets that collected in the droplet 15 well were then manually transferred with a multichannel pipet to a 96-well PCR plate. The plate was heat-sealed with a foil seal and then placed on a conventional thermal cycler and amplified to the end-point (40 cycles). Thermal cycling conditions were 95'C x 10 min (1 cycle), 94'C x 30 s and 57'C x 60 s (40 cycles), and 12'C hold. After PCR, the 96-well PCR plate was loaded on the droplet reader 20 (Bio-Rad), which automatically reads the droplets from each well of the plate. Analysis of the ddPCR data was performed with QuantaSoft analysis software (Bio Rad) that accompanied the droplet reader. Primer sequences: Forward: 5'GATAAAACTGAGCAAGAGGCTTTGG' (SEQ ID NO: 5), Reverse: 5'GCTGTTTAATTGTGTGGAAGATCCAA' (SEQ ID NO: 6), Wild type probe: 5' 25 VIC-CCACCATGATGTGCA-MGB-3 (SEQ ID NO: 7), 'H1047L probe:5'-FAM CCACCATGATGAGCA-MGB-3' (SEQ ID NO: 8), Hi 047R probe: 5'-FAM CCACCATGATGCGCA-MGB-3' (SEQ ID NO: 9). Subtyping TNBC cases and validating PIK3CA mutations in the TCGA. 30 RNA-seq (lvl 3), somatic mutations (lvl2) and reverse phase protein array (RPPA, lvl3) data for breast cancer was downloaded from The Cancer Genome WO 2013/075059 PCT/US2012/065724 - 61 Atlas (tcga-data.nci.nih.gov/). RNA-seq data (RPKM) was combined and molecularly subtyped using the online TNBCtype software according to published methods (Chen et al., Cancer Informatics, 11:147-1556 (2012)). After removal of potential ER positive, 137 TNBC samples were identified and assigned a molecular 5 subtype. Reverse Phase Protein Array on TNBC cell lines. Tumor or cell lysates were two-fold-serial diluted for 5 dilutions (from undiluted to 1:16 dilution) and arrayed on nitrocellulose-coated slide in 1 lx1I 10 format. Samples were probed with antibodies by CSA amplification approach and visualized by DAB colorimetric reaction. Slides were scanned on a flatbed scanner to produce 16-bit tiff image and the density was quantified by MicroVigene. Relative protein levels for each sample were determined by interpolation of each dilution curves from the standard curve (supercurve) of the slide (antibody) as 15 previously described (Zhang et al., Bioinformatics, 25:650-654 (2009)). These values were Log2 transformed and normalized for protein loading by transformation a to linear value. Linear normalized data was then median-centered for heatmap comparisons. 20 Cell proliferation/viability assays and IC 50 determinations. All cell lines were maintained in culture as previously described (JCI ref). Breast cancer cell lines and HMECs were seeded (3,000-10,000 cells) in quadruplicate wells in 96-well plates. Media was then replaced with either fresh media (control) or media containing half-log serial dilutions of the following drugs: 25 GDC-0941 (10-nM-3000nM), GDC0980 (1nM-300nM), BKM120 (10-nM 3000nM) and NVP-BEZ235 (lnM-300nM) purchased from Selleck Chemicals (Houston, TX) or in combination with bicalutamide (Sigma, St. Louis, MO). Viability was determined from measuring fluorescent intensity after metabolic reduction of Alamar Blue in the presence/absence of drug after 72 hours. Viability 30 assays were performed in triplicate and replicates were normalized to untreated wells. Inhibitory concentration (IC 50 ) values were determined after double-log WO 2013/075059 PCT/US2012/065724 - 62 transformation of dose response curves as previously described (Lehmann et al., J Clin Invest, 121:2750-2767 (2011)). Quantification of apoptosis 5 Caspase 3/7 activity Cell lines were seeded in triplicate in 96-well plates. Media was removed and replaced with media containing vehicle (control) or indicated drugs. After 48 hrs, apoptosis was determined by measuring luciferase from activated caspase 3/7 upon addition of Caspase-Glo reagent (Promega, Madison, WI). Relative levels of 10 caspase activity were normalized to viable cell number determined by metabolic reduction of Alamar blue. shRNA knockdown of AR 293FT cells were transfected with Lipofectamine 2000 (Invitrogen, Grand 15 Island, NY) the packaging vectors PAX2 and pMD2.g (Addgene, Cambridge, MA) along with pLKO. 1 -puro Misson shRNA constructs (Sigma) targeting AR or a non targeting control. Viral media was harvested 48h post-transfection and added to target cells along with polybrene (10 pg/mL). 72h post-infection MDA-MB-453 (8,000 cells/well), CAL-148 (5,000 cells/well) and MFM-223 (7,000 cells/well) 20 seeded in quadruplicate in 96-well plates. Following overnight attachment media was replaced with fresh media (control) or media containing half-log dilutions of drugs. Following incubation in drug for 72h viability was measured with alamarBlue as previously described (Lehmann et al., J Clin Invest, 121:2750-2767 (2011)). 25 Forced Suspension Viability Assay CAL-148, MFM-223 and MDA-MB-453 cells were plated (10,000 cells/well) in quadruplicate into 96-well plates coated with 0.9% agarose diluted in corresponding media. Following 5 d of growth, cells were then treated with 30 increasing half-log concentrations of GDC-0941 (10 - 3000 nM), GDC0980 (1 nM 300 nM), BKM120 (10 nM - 3000 nM) and NVP-BEZ235 (1 nM - 300 nM) alone WO 2013/075059 PCT/US2012/065724 - 63 or in combination with 25 iM bicalutamide (CDX) for an additional 72 h, upon which cells were imaged and viability determined by Alamar blue. Immunostaining 5 Formalin fixed paraffin embedded (FFPE) tissue were used for immunohistochemical studies. Androgen receptor (AR) expression and p-AKT expression were evaluated in FFPE tissue using the DAKO (Carpinteria, CA) antibody: AR (clone AR41 1) at a 1:50 dilution and AKT p473 (Cell signaling, CA) at a 1:200 dilution for 1 h at room temperature. FFPE tissue was subject to antigen 10 retrieval with high pH buffer (pH 8.0) followed by overnight incubation with an AR (1:30) or Ki67 (1:75) antibody dilution overnight. Immunoblotting Cells were trypsinized, lysed and relative protein expression was determined 15 by Western blot as previously described (JCI reference) with the following antibodies; the AR polyclonal antibody, SC-N20 (Santa Cruz Biotechnology), pS6 Ser235/236 (Cell Signaling #4856), S6 (Cell Signaling #2317), pAKT-S473 (Cell Signaling #9271), AKT (Cell Signaling #9272), PARP (Cell Signaling #9542) and GAPDH (Millipore). 20 Xenograft Tumor Studies. Five-week-old female athymic nude-Foxnlnu mice (Harlan Sprague Dawley) were injected (s.c.) with either -4 x 106 (CAL-148) or -10 x 106 (MDA MB-453) cells suspended in media (200 tL) into each flank using a 22-gauge 25 needle. Once tumors reached a volume of 25-50 mm 3 , mice were randomly allocated to treatment or vehicle arms. Treatments included bicalutamide p.o. (100 mg/kg/d), GDC0980 p.o. (7.5mg/kg/d) in 5% DMSO, 0.5% carboxymethyl cellulose (MCT) or NVP-BEZ235 p.o. (50 mg/kg/d) in 0.5% MCT or the combination of bicalutamide and either GDC-0980 or NVP-BEZ235 according to the above 30 concentrations. Tumor diameters were serially measured at the indicated times with digital calipers and tumor volumes were calculated as width 2 x length/2.

WO 2013/075059 PCT/US2012/065724 - 64 Supplemental Results Validation of digital droplet PIK3CA mutant PCR primers Primers were validated using on cell lines with known PIK3CA status cells. 5 The percentage of DNA reflected the frequency of mutant PIK3CA alleles similar to expected genotype status; HMECwt/wt = 0.01%, MDA-MB-453wt/mut = 61.19% and SUM-i 85mut/mut = 97.09% (Figure 26A). The slightly higher mutant frequency in the heterozygous mutant cell line MDA-MB-453 is likely caused by chromosomal aneuploidy. Digital droplet PCR was far more quantitative than 10 Sanger sequencing and both methods were 100% concordant when mutant DNA was present at levels greater than 5% (Figures 33A-33B). Example 3 Cell proliferation/viability assays and IC 50 determinations. All cell lines were 15 maintained in culture as previously described herein. Breast cancer cell lines and HMECs were seeded (3,000-10,000 cells) in quadruplicate wells in 96-well plates. Media was then replaced with either fresh media (control) or media containing half log serial dilutions of the following drugs: Sorafenib, OSi-906, BMS-754807 and PF2341066 purchased from Selleck Chemicals (Houston, TX) Viability was 20 determined from measuring fluorescent intensity after metabolic reduction of Alamar Blue in the presence/absence of drug after 72 hours. Viability assays were performed in triplicate and replicates were normalized to untreated wells. Inhibitory concentration (IC 5 ) values were determined after double-log transformation of dose response curves. 25 PDGFR inhibition selectively inhibits viability of mesenchymal-like TNBC cell lines. Bar graph displays the half-maximal inhibitory concentration (nM) of the PDGFR inhibitor sorafenib for a panel of TNBC cell lines. Colorbar below designates the TNBC subtype (basal-like= black, mesenchymal-like= grey) or normal cells (white). 30 IGF1R inhibition selectively inhibits viability of mesenchymal-like TNBC cell lines. Bar graph displays the half-maximal inhibitory concentration (nM) of the WO 2013/075059 PCT/US2012/065724 - 65 IGF1R inhibitor OSI-906 for a panel of TNBC cell lines. Colorbar below designates the TNBC subtype (basal-like= black, mesenchymal-like=grey) or normal cells (white). IGF1R inhibition selectively inhibits viability of mesenchymal-like TNBC 5 cell lines. Bar graph displays the half-maximal inhibitory concentration (nM) of the IGF1R inhibitor BMS-754807 for a panel of TNBC cell lines. Colorbar below designates the TNBC subtype (basal-like= black, mesenchymal-like= grey) or normal cells (white). MET inhibition selectively inhibits viability of mesenchymal-like TNBC cell 10 lines. Bar graph displays the half-maximal inhibitory concentration (nM) of the MET inhibitor PF2341066 for a panel of TNBC cell lines. Colorbar below designates the TNBC subtype (basal-like= black, mesenchymal-like= grey) or normal cells (white).

WO 2013/075059 PCT/US2012/065724 -66 N u - CA m u , 'C.. - l CN4 E 78 4,31 WO 2013/075059 PCT/US2012/065724 -67 - 5~ '-' -.. o ~ ~ E 8.9 0 EE A 0 0 0 0o 0 0 0 0 0 0 0 r WO 2013/075059 PCT/US2012/065724 -68 000 Co) o goC 0Tci~-o0 00 M 0CM 0 oD

CM

WO 2013/075059 PCT/US2012/065724 -69 co P 4, 0 4 4 ~0Cl -~)2 wL 0q 0 0 4. 9 ClCLb H0 Cl) WO 2013/075059 PCT/US2012/065724 -70 u 7 - 4 4 7 -C i -t~ ~ - - ~ -~ 043 0 00 0 C ~ CA - ~ ~ n 00 0cC 0 0W 00- C 0C e= 0 'C 0~ 0~ 0 0 0 0 0~ t V) C\ 0- c-) C)- C) w 0kl m 0, t.-~-'l 0 0 C Q u CA0 41 .0 WO 2013/075059 PCT/US2012/065724 -71 u Q 8 a "Z) ) C D iD m x kmU Ln ~ 0 H c.. Z m Ln c1J~ ~ u

~

C<ii2 Q u~ c ,0 4z EJfr c!

IN

WO 2013/075059 PCT/US2012/065724 -72 Table 4 Top gene ontologies for TNBC subtypes in the training and validation datasets Subtype Training Set (386) Validation Set (201) BLI DNA REPLCATION REACTOME CEL CYCLE KEGG CELL CYCLE KEGG DNA REPLICATION REACTOME RNA POLYMERASE G2 PATHWAY CELL CYCLE METHIONINE METABOLISM GLYCOSPHINGOLIPID BIOSYNTHESIS NEO CHOLESTEROL BIOSYNTHESIS LACTOSERIES CELL CYCLE ATRBRCAPATHWAY RNA POLYMERASE UBIQUITIN MEDIATED PROTEULYSIS ATRBRCAPATHWAY G2 PAT HWAY GI TO S CELL CYCLE REACTOME PROTEASOME CASPASE CASCADE G1 TO S CELL CYCLE REACTOME GLUTAMATE METABOLISM METHIONINE METABOUSM AMINOACYL TRNA BIOSYNTHESIS FASPATHWAY GLYCOSPHINGOLIPID BIOSYNTHESIS NEO LACTOSERIES CHOLESTEROL BIOSYNTHESIS 01 AND S PHASES PYRIMIDINE METABOLISM CYSTEINE METABOLISM AMINOACYL TRNA BIOSYNTHESIS ATMPATH1WAY DNA POLYMERASE MRNA PROCESSING REACTOME CARBON FIXATION BIOSYNTHESIS OF STEROIDS ALANINE AND ASPARTATE METABOLiSM KREBS TCA CYCLE PROTEASOME PYRIMIDINE METABOLISM CELLOYCLEPATHWAY PROTEASOMEPATHNWAY CDMACPATHWAY SELENOAMINO ACID METABOLISM KREBS TCA CYCLE CELLCYCLEPATHWAY ALANINE AND ASPARTATE METABOUSM GAO PATHWAY AMINOACY L'TRNA BIOSYNTHESIS FASPATHWAY

PROTEASOME

WO 2013/075059 PCT/US2012/065724 -73 Table 4 Top gene ontologies for TNBC subtypes in the training and validation datasets Subtype Training Set (386) Validation Set (201) BL2 PORPHYRIN AND CHLOROPHYLL METABOLISM BLADDER CANCER GATA3PATHWAY GATA3PATHWAY STEMPATHWAY EPITHELIAL CELL SIGNALING IN HELICOBACTER PYLORI PARKINSONS DISEASE INFECTION STARCH AND SUCROSE METABOLISM CARDIACEGFPATHWAY EGFPATH WAY NICOTINATE AND NICOTINAMIDE METABOLISM INSULINPATHWAY EDoG1PATHWAY RENAL CELL CARCINOMA PANTOTHENATE AND COA BIOSYNTHESIS WNT SIGNALING TOLLPATHWAY AMINOACYL TRNA BIOSYNTHESIS RENLN ANGIOTENSN SYSTEM GALACTOSE METABOLISM ECM RECEPTOR INTERACTION O GLYCAN BIOSYNTHESIS PHO TOPYTHE ATMPATHWAY PHOTOSYNTHESIS 1- W NTINYTONS DISEASE NTHPATEALIAY ILB1PATH'AWAY AINOSUOARS METABOLISM 01 TO S CELL CYCLE REACTOME EICOSANOID SYNTHESIS FOCAL ADHESION DORSO VENTRAL AXIS FORMATION IGF1PATHWAY NGFPATHWAY UBlQUITIN MEDIATED PROTEOLYSIS SPRYPATHWAY GLOA ECMPATHWAY TPDPATHWAY METPATHWAY G2PATHWAY CDMACPATH WAY WNT BETA CATENIN PATHWAY ARFPATH WAY INSUIN RECEPTOR PATHWAY IN CARDIAC P38 MAPK PATHWAY MYCOYTES CHOLERA INFECTION METPATHWAY HYPERTROPHY MODEL P53 SIGNALING PATHWAY WNT BETA CATENIN PATHWAY CARDIACEGFPATWAY ERKPATHWAY RASPATH W'AY NFKBPATHWAY FMLPPATHWAY UCALPAINPATHWAY GLYCOLYSIS GSK3PATHWAY ATMPATHWAY GLYCOLYSIS AND GLUCONEOGENESis IGF1RPATHWAY EPOPATHWAY GLYCOSAMINOGLYCAN DEGRADATION SPRYPATHWAY GALACTOSE METABOLISM TRKA RECEPTOR ERKI ERK2 MAPK PATHWAY PHOSPHOINOSITIDE 3 KINASE PATHWAY GLUCONEOGENESIS _ PROTEASOME I AND 2 METHYLNAPHTH-ALENE DEGRADATION P38 MAPK PATHWAY VEGFPATH WAY PATHOGENIC ESCHERICHIA COU INFECTON EPEC CHOLERA INFECTION ETSPATHWAY TIDPATHWAY PROSTAGLANDIN SYNTHESIS REGULATION ECMPATHWAY BENZOATE DEGRADATION VIA COA LIGATION POGFPATHWAY NFIKBPATHWAY ARGININE AND PROL[NE METABOLISM HEPARAN SULFATE BIOSYNTHESIS GLYCOLYSIS PROTEASOMEPATHWAY NTHIPATHWAY EPITHELIAL CELL SIGNALING IN HELICOBACTER PYLORI INFECTION NOPPATHWAY ETS PATHWAY ERKI ERK2 MAPK PATHWAY GALACTOSE METABOLISM GAP JUNCTION ERYTHPATHWAY GLYCOLYSIS AND GLUCONEOGENESIS MYOCYTE AD PATHWAY WO 2013/075059 PCT/US2012/065724 -74 Table 4 Top gene ontologies for TNBC subtypes in the training and validation datasets Subtype Training Set (386) Validation Set (201) IM CTLA4PATHWAY NO2IL12PATHWAY NO 22PATHWAY CTLA4PATHAY THiTH2PATHWAY IIP~lA CSKPATHWAY L1PATWAY NKTPATHWAY TYPE I DIABETES 0ELLITUS COMPPATHWAY NKCELLSPATHWAY LPRTHAWE THITH2PATHWAY NKCELLSPATHWAY TNFR2PATHWAY TYPE I DIABETES MELLITUS CSKPATHWAY AMIPATHWAY IL7PATHWAY ANTIGEN PROCESSING AND PRESENTATION AMIPATHWAY IL112PATH WAY LAIRPATHHWY ANTIGEN PROCESSING AND PRESENTATION TNFR2PATH WAY NFKBPATHWAY DCPATHWAY TUMOR NECROSIS FACTOR PATHWAY TIDPATHWAY PROTEgASOg E PTOS IS GENMAPP NKTPATHWAY NFKBPATH WAY PROTEASOMEPATHWAY T CELL SIGNAL TRANSDUCTION T CELL SIGNAL TRANSDUCTION DEATHPATHWAY TUMOR NECROSIS FACTOR PATHWAY APOPTOSIS GENMAPP CASPASEPATHWAY APOPTOSIS CAH Y NATURAL KILLER CELL MEDIATED CYTOTOXICITY RELAPATHWAY TOBiPATHWAY T081PATHWAY B CELL RECEPTOR SIGNALING PATHWAY APOPTOSIB T CELL RECEPTOR SIGNALING PATHWAY APOPTOSIS KEGG CYTOKINEPATH4WAY CELL ADHESION MOLECULES T CELL RECEPTOR SIGNALING PATHWAY DEATHPATHWAY CYTOKINEPATIHWAY APOPTOSIS KEGO CASPASEPATHWAY TOLL LIKE RECEPTOR SIGNALING PATHWAY TIDPATHWAY PROTEASOME HIEMATOPOIETIC CELL LINEAGE IL2PATHwAy RELAPATHWAY NATURAL KILLER CELL MEDIATED CYTOTOXICITY IL2PATHWAY CASPASE CASCADE MITOCHONDRIAPATHWAY LAIRPATHWAY TOLLPATHWAY B CELL ANTIGEN RECEPTOR MIEMATOPOIETIC CELL LINEAGE INFLAMPATHWAY ILPATHWAY IL2RBPATHWAY COMPPATHWAY ILSPATHWAY TRAHA STEMPATH WAY BCR SIGNALING PATHWAY B CELL RECEPTOR SIGNALNG PATHWAY HIVNEFPATHWAY STEMPATHWAY TCRPATHWAY M[TOCHONDRIAPATHWAY CASPASE CASCAIDECELA EIOMLCLS PIP3 SIGNALING N14 B LYMPHOCYTES CELL ADHESION OLECULES NTHIPATHWAY 41RBBPATHWAY STRESSIPATHWAY CCR51PATHWAY IL2RSPATHWAY CYTOKINE CYTOKINE RECEPTOR INTERACTION INFLAMPATHWAY INTERLEUKIN 4 PAT14WAY CYTOKINE CYTOKINE RECEPTOR INTERACTION STRESSPATHWAY PIP3 SIGNALING IN B LYMPHOCYTES IL1RPATHWAY TOLL UKE RECEPTOR 81GNALING PATHWAY APOPTOSIS 19AH Y EICOSANOID SYNTHESIS UBIQUITIN MEDIATED PROTEOLYSIS B CELL ANTIGEN RECEPTOR CELLCYCLEPATH WAY 8CR StGNALING PATHWAY CALCINEURIN NF AT SIGNALING PROTEASOME FC EPSILON RI SIGNALING PATHWAY CERAMIDEPATHWAY AMINOACYL TRNA BIOSYNTHESIS APOPTOSIS JAK STAT SIGNALING PATHWAY JAK STAT SIGNALING PATHWAY FASPATHWAY EPOPATHWAY PTEN PATHWAY NTHIPATH WAY WO 2013/075059 PCT/US2012/065724 -75 Table 4 Top gene ontologies for TNBC subtypes in the training and validation datasets Subtype Training Set (386) Validation Set (201) IM B CELL RECEPTOR COMPLEXES CD40PATHWAYMAP cont'd DNA REPLICATION REACTOME CALCINEURIN NF AT SIGNALING FOLATE BIOSYNTHESIS HIVNEFPATHWAY ATRBRPATHWAY EICOSANOID SYNTHESIS LEUKOCYTE TRANSENDOTHELIAL MIGRATION BCRPATHWAY GLEEVECPATHWAY FC EPSILON RI SIGNAUNG PATHWAY TNFRIPATHWAY KERATINOCYTEPATHWAY COMPLEMENT AND COAGULATION CASCADES AMINOACYL TRNA BIOSYNTHESIS CRPATHWAY CELLCYCLEPATHWAY GHPATHWAY B CELL RECEPTOR COMPLEXES TPOPATHWAY IL6PATHWAY AMINOACYL TRNA BIOSYNTHESIS PROTEASOMEPATH WAY PEPTIDE GPCRS CCRSPATHWAY N GLYCAN BIOSYNTHESIS ATRBRCAPATHWAY ALZHEIMERS DISEASE ACUTE MYELOID LEUKEMIA GHPATHWAY GAQ PATHWAY FOLATE BIOSYNTHESIS CD40PATHWAYMAP PEPTIDE GPCRS METHIONINE METABOLISM GLEEVEC PATHWAY GPORDB CLASS B SECRETIN LIKE PASPATHWAY CRCDEPTHN B LYPHOCYTES ABC TRANSPORTERS GENERAL FCERIPATHWAY INTERLEUKIN 4 PATHWAY G1 TO S CELL CYCLE REACTOME IL4RECEPTOR IN B LYPHOCYTES PARKINSONS DISEASE ADIPOCYTOKINE SIGNALING PATHWAY CHEMICALPATHWAY PROSTAGLANDIN AND LEUKOTRIENE METABOLISM CELL CYCLE KEO0G SNARE INTER ACTIONS IN VESICULAR TRANSPORT CDMACPATH WAY ATMPATHWAY TPOPATHWAY NICOTINATE AND NICOTINAMIDE METABOLISM HSP27PATHWAY KERATINOCYTEPATHWAY TNER1PATHWAY HYPERTROPHY MODEL (NOSITOL PHOSPHATE METABOLISM ADIPOCYTOKINE SIGNALING PATHWAY DENTATORUBROPALUDOLUYSIAN ATROPHY ACUTE MYELOID LEUKEMIA FMLPPATHWAY LEUKOCYTE TRANSENDOTHELIAL MIGRATION DICTYOSTELIUM DISCOIDEUM CAMP CHEMOTAXIS ILIRPA THWAY PATHWAY DENTATORUBROPALLIDOLUYSIAN ATROPHY BIOPEPTIDESPATHWAY OVARIAN INFERTILITY GENES PANCREATIC CANCER FAS SIGNALNG PATHWAY PTEN PATHWAY RACCYCDPATHWAY AMINOSUGARS METABOLISM WO 2013/075059 PCT/US2012/065724 -76 Table 4 Top gene ontologies for TNBC subtypes in the training and validation datasets Subtype Training Set (386) Validation Set (201) M CELL COMM. UNICATION PTENPATHWAY HEPARAN SULFATE BIOSYNTHESIS DNA REPLCATION REACTOdE ECM RECEPTOR INTERACTION RIBOSOME ALKPATHWAY TRANSLATION FACTORS UCALPAINPATHWAY RNA POLYMERASE STRIATED MUSCLE CONTRACTION DNA POLYMERASE BASAL CELL CARCINOMA RAAL. TRANSCRIPTION FACTORS REGULATION OF THE ACTIN CYTOSKELETON BY IGFIMTORPAT HWAY RHO GTPASES RNA TRANSCRIPTION REACTOME TGF BETA SIGNALING PATHWAY CHOLESTEROL BIOSYNTHESIS H AHPATH WAY GLyLPCOSYLP 1OSPHATIDYLINOSITOL ANCHOR BIOSYNTHESIS HEDGEHOG SIGNALING PATHWAY ECMPATH'WAY INOSITOL PHOSPHATE METABOLISM UCALPAINPATHWAY IGF1PATHWAY EIF4PATHWAY RKPATWY RIBOSOMAL PROTEINS FOCALADHESION CELL CYCLE KEGG EDGiPAAHWAY MTORPATHWAY CARM ERPATHWAY REGULATION OF THE ACTIN CYTOSKELETON BY RHO INSULINPATHWAY GTPASES ETHER LIPID METABOLISM G1 TO S CELL CYCLE REACTOME TELPATHWAY MRNA PROCESSING REACTOME NOTCH SIGNALING PATHWAY USIQUITIN MEDIATED PROTEOLYSIS WNT BETA CATENIN PATHWAY WNTPATHWAY MEF2DPATH WAY ALKPATHWAY INTEGRINPATHWAY G2PATHWAY MELANOGENESIS TELPATHWAY VEGFPATIWAY CELL CYCLE WNTPATHWAY ATRBRCAPATHWAY ETSPATW AY DORSO VENTRAL AXIS FORMATION WO 2013/075059 PCT/US2012/065724 -77 Table 4 Top gene ontologies for TNBC subtypes in the training and validation datasets Subtype Training Set (386) Validation Set (201) MSL -ROSTAGLANDIN SYNTHESIS REGULATION NO2IL2PATHWAY BADPATHWAY CSKPATHWAY LAIRPATHWAY LAIRPATHWAY ILTPATKWAY TOBiPATHWAY COMPPATHWAY CTLA4PATHWAY RENIN ANGIOTENSIN SYSTEM AMIPATHWAY AMIPATHWAY THiTH2PATHWAY CS KPATH WAY RENIN ANGIOTENSIN SYSTEM ERYTHPATHWAY PARIPATHWAY ECM RECEPTOR INTERACTION NKTPATHWAY TOBIPATHWAY IL7PATHWAY COMPLEMENT AND COAGULATION CASCADES NKCELLSPATHWAY CARDIACEGFPATHWAY HISTIDINE METABOLISM ECPTHWAY GATA3PAIHWAY ERYTHPATH WAY NDKDYNAMINPATHWAY HEMATOPOIETIC CELL LINEAGE CCRSPATHWAY TYPE I DIABETES MELUTUS FOCAL ADHESION BETA ALANINE METABOLISM GLYCOSAMINOGLYCAN DEGRADATION INTRINSICPATHWAY IL12PATHWAY PARIPATHWAY ItL2PATHWAY ALKPATHWAY PROSTAGLANDIN SYNTHESIS REGULATION CALCINEURINPATHWAY EicOSANOID SYNTHESIS NTHIPATHWAY COMPPATHWAY CTLA4PATHWAY CCRSPATHWAY BETA ALANINE METABOLISM GLYCAN STRUCTURES DEGRADATION GCRPATHWAY CALCINEURINPATHWAY HEMATOPOIETIC CELL LINEAGE COMPLEMENT AND ODAGULATION CASCADES NO1PATHWAY ECM RECEPTOR INTERACTION VIPPATH WAY TCRPATH WAY UCALPAINPATHWAY TNFR2PATH WAY EDGI PATHWAY NOiPATHWAY OlPATHWAY INTRINSICPATHWAY IGF PTHWAY TCRPATHW.IAY STAIN PATHWAY PHARMGX0 EICOSANOID SYNTHESIS ALKALOID BIOSYNTHESIS 11 NKCELLSPATHWAY LEUKOCYTE TRANSENDOTHELIAL MGRATION, VALINE LEUCINE AND ISOLEUCINE DEGRADATION HSA04S14 CELL ADHESION MOLECULES PPARAPATHWAY jNFLAMPAThWAY SPRYPATHWAY MEF2DPATHWAY HDACPATHWAY 1L2RBPATHWAY STATIN PATHWAY PHARMGKB WNT BETA CATENIN PATHWAY GLYCAN STRUCTURES DEGRADATION HSA04510 FOCAL ADHESION SMOOTH MUSCLE CONTRACTION NTHIPATHWAY THiTH2PATHWAY BCRPATHWAY ECMPATHWAY ,8CR SIGNALING PATHWAY GPCRPATHWAY GCRPATH WAY PDGFPATHWAY GLYEROLIPID IETABOISM INTEGRIN MEDIATED CELL ADHESION KEGO HISTIDINE METABOLISM FATTY ACID METABOLISM GATA3PATHWAY NKTPATH WAY PEPTIDE GPCRS CXCR4PATH WAY PIP3 SiGNALING IN B LYMPH-OCYTES TOLLPATHWAY CYTOKINE CYTOKINE RECEPTOR INTERACTION TNFR2PATHWAY INOSITOL PHOSPHATE METABOLISM LEUKOCYTE TRANSENOOTHELIAL MIGRATION BADPATHWAY PPAR SIGNALING PATHWAY MONOAMINE GPORS AMYOTROPHiC LATERAL SCLEROSIS NFKBPATHWAY GLYCEROLIPID METABOLISM ECMYPATHWAY GLYCOSPHINGOLIP1D METABOLISM i AND 2 METHYLNAPHTHAL ENE DEGRADATION MTORPATHWAY B CELL RECEPTOR SIGNALING PATHWAY NICOTINATE AND NiCOTINAMIDE METABOLISM VIPPATHWAY PGC1APATHWAY CARDIACEGFPATHWAY METPATHWAY ALKPATHWAY WNT BETA CATENIN PATHWAY IL4RECEPTOR IN B LYPHOCYTES TPOPATHWAY GHPATHWAY BUTANOATE METABOLISM NATURAL KILLER CELL MEDIATED CYTOTOXICITY WO 2013/075059 PCT/US2012/065724 -78 Table 4 Top gene ontologies for TNBC subtypes in the training and validation datasets Subtype Training Set (386) Validation Set (201) MSL TGF BETA SIGNALING PATHWAY RACIPATHWAY con'd GSK3PATHWAY EDGIPATHWAY GLYCOSAMINOGLYCAN DEGRADATION BETA ALANINE METABOLISM PROPANOATE METABOLISM UCALPAINPATHWAY CELL COMMUNICATION OVARIAN INFERTILITY GENES GPCRPATHWAY VALINE LEUCINE AND ISOLEUCINE DEGRADATION ERK1 ERK2 MAPK PATHWAY BILE ACID BIOSYNTHESIS INTEGRIN MEDIATED CELL ADHESION KEGG ILBPATHWAY ARACHIDONIC ACID METABOLISM OCR SIGNALING PATHWAY ABC TRANSPORTERS GENERAL CELL ADHESION MOLECULES RHOPATHWAy B CELL RECEPTOR SIGNALING PATHWAY R A LPATH WAY BCRPATHWAY JAK STAT SIGNAING PATHWAY PTENPATHWAY HDACPATh WAY MCALPAINPATHWAY BILE ACID BIOSYNTHESIS 1 AND 2 METHYLNAPHTHALENE DEGRADATION TOLLPATHWAY GPCRDB CLASS B SECRETIN LIKE SPPAPATH WAY IL2PATHWAY NDKDYNAMIN PATHWAY PROPANOATE METABOLISM SMOOTH MUSCLE CONTRACTION MELANOMA INTEGRINPATHWAY CALCIUM SIGNALING PATHWAY ALKALOID BIOSYNTHESIS || GLYCEROLIPID ME T ABOUSM ETHER LIPID METABOLISM ANTIGEN PROCESSING AND PRESENTATION NUCLEAR RECEPTORS ADIPOCYTOKINE SIGNALING PATHWAY CKIPATHWAY TYPE I DIABETES MELLITUS NO2Li2PATHWAY RHOPATHWAY CYTOKINE CYTOKINE RECEPTOR INTERACTION BLOOD CLOTTING CASCADE INFLAMPATHWAY DICTYOSTELIUM DISCOIDEUM CAMP CHEMOTAXIS PATHWAY DIFFERENTIATION PATHWAY IN PCI CELLS WO 2013/075059 PCT/US2012/065724 -79 Table 4 Top gene ontologies for TNBC subtypes in the training and validation datasets Subtype Training Set (386) Validation Set (201) LAR CHOLESTEROL BIOSYNTHESIS GLU TATHIONE METABOLISM PENTOSE AND GLUCURONATE PENTOSE AND GLUCURONATE INTERCONVERSIONS INTERCONVERSIONS GLUTATHIONE METABOLISM BIOSYNTHESIS OF STEROIDS CHOLESTEROL BIOSYNTHESIS TYROSINE METABOLISM TYROSINE METABOLISM GAMMA HEXACHLOROCYCLOHEXANE BIOSYNTHESIS OF STEROIDS DEGRADATION PORPHYRIN AND CHLOROPHYLL METABOLISM PORPHYRIN AND CHLOROPHYLL METABOLISM ANDROGEN AND ESTROGEN METABOLISM PHENYLALANINE METABOLISM GLYCOSPHINGOLIPID METABOLISM GAMMA HEXACHLOROCYCLOHEXANE TYPE III SECRETION SYSTEM DEGRADATION GAMMA HEXACHLOROCYCLOREXANE DEGRADATION CHREBPPATHWAY FLAGELLAR ASSEMBLY GLIITATHIONE METABOLISM CITRATE CYCLE TCA CYCLE I AND 2 METHYLNAPHTHALENE DEGRADATION PHENYLALANINE METABOLISM OLYCOSPHINGOLIPID METABOLISM ATP SYNTHESIS PORPHYRIN AND CHLOROPHYLL METABOLISM -PHOTOSYNTHESIS ENAND ESTROGEN METABOL-ISM SAC N URS EAOIS PiNnrLALANINE METABOUSM PORPHYRIN AND CHLOROPHYLL METABOLISM CITRATE CYCLE TCA CYCLE ARGININE AND PROLINE METABOLISM TYROSINE METABOLISM GAMMA HEXACHLOROCYCLOHEXANE DEGRADATION METABOLISM OF XENOBIOTICS BY CYTOCHROME METABOLISM OF XENOBIOTICS BY CYTOCHROME P450 P450 FRUCTOSE AND MANNOSE METABOLISM EICOSANOID SYNTHESIS CARBON FIXATION VALINE LEUCINE AND ISOLEUCINE DEGRADATION GLYCAN STRUCTURES DEGRADATION GLYCAN STRUCTURES DEGRADATION CITRATE CYCLE GLYCOSYLPHOSPHATIDYLINOSITOL ANCHOR COMPPATHWAY BIOSYNTHESIS PARKINSONS DISEASE BIJTANOATE METABOLISM ANDROGEN AND ESTROGEN METABOLISM ARGININE AND PROLINE METABOLISM PANTOTHENATE AND COA BIOSYNTHESIS GLUTATHIONE METABOLISM FATTYACID METABOLISM CKIPATHWAY CITRATE CYCLE REIN ANGIOTENSIN SYSTEM PANTOTHENATE AND COA BIOSYNTHESIS TYROSINE METABOLISM PROPANOATE METABOLISM GLUTAMATE METABOLISM FATTY ACID METABOLISM ALANINE AND ASPARTATE METABOLISM ANDROGEN AND ESTROGEN METABOLISM CARBON FIXATION GLUTAMATE METABOLISM EICOSANOD SYNTHESIS N GLYCAN BIOSYNTHESIS ECSNI YTEI HISTINE METAB SOLISM GLYCOSAMINOGLYCAN DEGRADATION BILE ACID BIOSYNTHESIS PHENYLALANINE METABOLISM PROPANOATE METABOLISM CHREBPPATHWAY HCMVPATHWAY CKiPATHWAY TRYPTOPHAN METABOUSM MONOAMINE GPCRS ALANINE AND ASPARTATE METABOLISM SPHINGOLIPID METABOLISM VALNE LEUPINE AND ISOLEUCINE DEGRADATION P53HYPOXIAPATHWAY PPAR SIGNALING PATHWAY TRYPTOPHAN METABOLISM CRESPATH WAY NICOTINATE AND NICQTINAMIDE METABOLISM ATP SYNTHESIS GLUTAMATE METABOLISM FLAGELLAR ASSEMBLY UCALPAINPATHWOIAY PE I SECRETION SYSTEM FRUCTOSE AND MANNOSE METABOLISM ERYTHPATHWAY KREBS TCA CYCLE PHOTOSYNTHESIS OXIDATIVE PHOSPHORYLATION HISTIDINE METABOLISM HISTIDINE METABOLISM MCALPAINPATHJWAY GALACTOSE METABOLISM FRUCTOSE AND MANNOSE METABOLISM PENTOSE PHOSPHATE PATHWAY GLYCOSAMINOGLYCAN DEGRADATION OXIDATIVE PHOSPHORYLATION SPHINGOLIPID METABOLISM N GLYCAN BIOSYNTHESIS GLUTAMATE METABOLISM UNOLEIC ACID METABOLISM LINOLEIC ACID METABOLISM BLOOD CLOTTING CASCADE GLYCEROLIPID METABOLISM ARACHIDONIC ACID METABOLISM ARGININE AND PROLINE METABOLISM MCALPAINPATHWAY SNARE INTERACTIONS IN VESICULAR TRANSPORT ERYTHPATHWAY SELENOAMINO ACID METABOLISM ABC TRANS PORT RS GENERAL PENTOSE PHOSPHATE PATHWAY FRUCTOSE AND MANNOSE METABOLISM EUTANOATE METABOLISM 1, NOIPATHWAY 0 GLYCAN BIOSYNTHES IS BILE ACID BIOSYNTHESIS PPAR SIGNALING PATHWAY ALANINE AND ASPARTATE METABOLISM VALINE LEUCINE AND ISOLEUCINE DEGRADATION WO 2013/075059 PCT/US2012/065724 -80 Table 4 Top gene ontologies for TNBC subtypes in the training and validation datasets Subtype Training Set (386) Validation Set (201) LAR cont'd STARCH AND SUCROSE METABOLISM NO1 PATHWAY YSTE TABOISM STARCH AND SUCROSE METABOLISM ALANINE AND ASPARTATE METABOLISM ARGININE AND PRALINE METABOLISM LIMONENE AND PINENE DEGRADATION COMPLEMENT AND COAGULATION CASCADES ABC TRANSPORTERS GENERAL GALACTOSE METABOLISM UREA CYCLE AND METABOLISM OF AMINO PENTOSE PHOSPHATE PATHWAY GROUPS OXIDATIVE PHOSPHORYLATION N GLYCAN BIOSYNTHESIS GLYCEROLIPID METABOLISM KREBS TCA CYCLE AMINOSUGARS METABOLISM ARACHDONIC ACID METABOLISM TRYPTOPHAN METABOLISM PENTOSE PHOSPHATE PATHWAY NAPHTHALENE AND ANTHRACFNE DEGRADATION OXIDATIVE PHOSPHORYLATION UNC DNA REPUCATION REACTOME ATRBRCAPATHWAY DNA POLYMERASE CELL CYCLE KEGG CELLCYCLEPATHWAY 01 TO S CELL CYCLE REACTOME CELL CYCLE 01PATHWAY 01 AND S PHASES BASAL TRANSCRIPTION FACTORS RNA TRANSCRIPTION REACTOME AMINOACYL TRNA BIOSYNTHESIS RNA POLYMERASE PYRIMIDINE METABOLISM CARM ERPATH'WAY PYRIMIDINE METABOLISM ALANINE AND ASPARTATE METABOLISM ALANINE AND ASPARTATE METABOLISM AMINOACYL TRNA BIOSYNTHESIS G2PATHWAY P53PAT HWAY SELENOAMINO ACID METABOL ISM GLUTAMATE METABOLISM PENTOSE PHOSPHATE PATHWAY MRNA PROCESSING REACTOME PENTOSE PHOSPHATE PATHWAY Bold- shared between tralmIng and validation set WO 2013/075059 PCT/US2012/065724 C) 0 CD cD -- t toOO O ' -I u -u - u -u - u u uu u uu u VJ VJ)CCCC ) ) C N CC c C ) C C C D uC C)))kn ))))C ) C )))C -u Q~ -~~0 -. 000 -C CCO ~~~~ Ln)) ) ) C C)o C) u) C WO 2013/075059 PCT/US2012/065724 -82 cn + + C/ V) CIOl cn C) n c ++ c -d u UQ U pq H H u0 0u - 0q o,, CI -~~QQQQ U WO 2013/075059 PCT/US2012/065724 -83 Table 6A Gene Categories BLI BL2 IM M MSL LAR ACTG2 ALOX5 ADAMDEC1 ACTA2 ABCA81 ABCA12 ADAMDEC1 CLCA2 AIM2 ACTG2 ABCB1 AKAP12 ANP32E COL5A1 ALOX5 BCL11A ACTA2 AKR1B1O APOBEC3A CTSC ANP~32E BGNE ACTG2 AKR1DI ATR3 CXCL12 APOBEC3A CAV I F AKAP12 ALCAMG AURKA A DHCR24 APOBEC3G CAV2F ALDHAIL ALDH3B2" AURKB A EGFRD BCL2AI CCND2' ALOX5 ALOX15B AZGP1 EPHA2 BIRC3 COL3A1 AOC3 APOD BCLlIA GALNIO BTK' COL5A1E APOBEC3G ARE' BIRC5 GBP2 BTN3A3 COL6A3 E AZGP1 BLVRB BUB I A GPR87 CIQA COL6A3E AZGP1 BLVRB CCL5 HTRA1 C1QB CTNNB I F BCL2 C413 D CADPF CCNA2 MET C413 DKK2 BGN"LCDS CENPA A MME CASPI DklO BMP2 ECLCA2 CENPF A PYCARD CCL 19H ENI C413 CLDN8 CHEKIc RSAD2 CCL3 HEPHB3 CV 1 COL5AI CKAP2 S100A7 CCL4H FBN1 £ CAV2V COL6A3 COBL S100A8 CCL5H FN1 F CCND2F- CRAT CTSS S100A9 CCL8H FOXCI CCR211 CUX2 CXCLIO TNFSF1O CCR1 HZ CD2DHR4 CXCLI 1 TP63 CCR2 H GNG1 1 CD37 DHRS2 CXCL13 CCR5 H GPR161 CD3D EAF2 CXCL9 CCR7 H HTRAI CD48 FAR ENI CD2K GI CD52 FASN EPH133 CD37K GF CD69 FBN1 EXOl c CD3 8K IGFBP4 CD74 FKBP5G FANCAc CD3D KJGFBP5 CKAP2 FMO5 FANCOc CD48 KKCNK5 COL3A1£ FOXAl"; GBP1 CD52 LAMB 1 COL5A1 F G6PD IF144L CD69K MFGE8 COL5A27E CALNT1O IFIHI CD74 K MIA COL6A3 - GALNT7 JGKC CD8A K MICALLI CORONA GGT1 IGKVID-13 CKAP2 TM 2CSF2RB GGTLC3 WO 2013/075059 PCT/US2012/065724 -84 Table 6A Gene Categories BLI BL2 TM M MSL LAR IGKV4-1 COROlA NOCIE CTNNB1 F HGD JTPKB CSF2RB PDGFRA CXCL12 11 f HMGCS2 IVD CTSC PDGFRB CXCL13 Hff HSD17B1 1 LOCI00130100 CTSS PMP22 CYTIP HTRAI L0C339562 CXCLIOH ROPNI DKK2F INPP4B L0C652493 CXCL1 1 H S100A1 DKK3 FIQGAP2 MCMIOA CXCL13 H SCRG1 DPYD KCNMA1 MDC1 c CXCL14 H SFR 'ENG KMO MFGE8 CXCL9 H SA6EPASI KRT1 8u MIA CXCR4H SMAD7~ LV2 KYNU MICALLI CXCR6 H SNA12 "EV12B LRRC17 MSH2 cCYBB SOXIO FBN1' MLPH MUC5B3 CYTIP SPARC FCFL2 MSX2 MYBLI DDX60 SPOCKI FGL2 pip u MYC 1 DPYD SRPX FNI E POU2AFI NBN u ENI SYNM FYB RARRES3 NRAS' EV12A TAGLN FZD4 S SI 00A7 NTN3 EV12B TCF4 " GALNT1O S I00A8 PAD12 FGL2 TCF7L2F GALNT7 S100A9 PLK1 A FYB TGFB1 E GIMAP6 SERHL2 PMP22 GBP1 TGFB1LI' GNGI I SIDTI POU2AFI GBP2 TGFB2 E GZMA SPARCLI PRCI A GIMAP6 TGFB3E HCLSI SPDEFG PSMB9 GNLY TGFBRI " HLA-DMA SPOCKI RAD21IA GZMA TGFBR2" H-LA-DPB1 TARP RAD5 1 cGZMB TGFBR3 E' HLA-DRA TFAP2B RAD54BP c HCLS1 TRPS1 HLA-DRB1I TNFSF1O ROPNl HCP5 TWIST I E LOXAIOMFIS TP53TGI RSAD2 HERC5 vim 1 HOXCA5 TRGC2 RTP4 HLA-DMA WWP2 HSD1I7B 11 TRGV9 SlOGAl HLA-DPB1I ZEB I HTRAI TRIM36 SCRG1 HLA-DRA ZEB IGFI1 1 UGT2B328 SOXIO HLA-DRB1 IGF2 E XBPI' WO 2013/075059 PCT/US2012/065724 -85 Table 6A Gene Categories BLI BL2 IM M MSL LAR SYNM IDOl IGFBP4 ZBTB 16 TAPI IF135 IGFBP5 TCF7L1 IF144L JGKC TP53BP2u IFIHI IGKVID-13 TRPS1 IFNG' IGKV4-1 TTK B IGKC IL23A UBD B IGKVID-13 IL2RG IGKV4-1 JQGAP2 ILIORAJ IRF8 IL15' ITGAV IL15RA' ITPKB IL16' IVD IL18' KDR IL18R1' KDR IL18RA' LAMBI IL IRN' LCK 1L23A' LHFP IL2RA' LOCI00130100 IL2RB' L0C339562 IL2RG ' L0C652493 IL2RG J LRRC 17 IL4' LTB 1L6' EI2

IL

7 ' MEOX IL IL7R ' MEOX21L IRFI' MMP2 Jpjp7' MSX1 IRF8' NGFR JTK NOTCH IVD NT5E JAKI NTN3 JAK2 OGN KYNU PDGFD WO 2013/075059 PCT/US2012/065724 -86 Table 6A Gene Categories BUi BL2 IM M MSL LAR LAMP3 PDGFRA LCK PDGFRB LOC100130100 PERI L0C339562 PLAC8 L0C652493 PMP22 LTB POU2AF1 LYN' PROCRL NFKBI' PTPRC NKG7 RAC2 NTN3 ROPN1 GASI SAMSN1 OAS2 SELL OASL SFPR4F PLAC8 SMAD6E POU2AF1 SMAD7 1 PSMB9 SNA12 1 PTPRC SPARE PYCARD SPARCL1 1 E RAC2' SPOCK1 RARRES3 SRGN RELB SRPX RSAD2 SYNM RTP4 TAGLNE S I00A8 TCF4F SAMSN1 TCF7L2 F SELL TEK SLAMF7 TERF21P SNX1O TGFBIE SRGN TGFBILI E STAWl TGFB2 E STAT4' TGFB3 E STAT5A' TGFBRI1E TAPI TGFBR2 E WO 2013/075059 PCT/US2012/065724 -87 Table 6A Gene Categories BLI BL2 IM M MSL LAR TARP TGFBR3 TFEC THBS4 TNFAIP2 THY1L TNFRSF17 TIEl TNFSF1O TNFRSF17 TRAC TRAC TRGC2 TRIM22 TRGV9 TWIST1I TRIM22 VCAM1 UBD VIME UBE2L6 XBP1 WARS ZBTB16 ZEBI ZEB23 A Cell division AR responsive 'Cell proliferation H Chemokine c DNA damage 'Immune signal transduction D Growth factor J Cytokine signaling E ECM/TGFB K Surface antigen Y WNT L Mesenchymal stem genes WO 2013/075059 PCT/US2012/065724 -88 Table 6B TNBC Subty e Mutations BLI BL2 IM M MSL LAR TOP2B BRCA1 TP53 APC APC MAP3K1 ATM BRCA2 STAT4 BRAF BRAF PIK3CA ATR CDKN2A STATIC CTNNB1 BRCA RB1 BRCA1 PTCH1 RET FGFR1 CDKN2A TP53 BRCA2 PTCH2 NFKBIA GLIl CTNNB1 CAMK1G PTEN MAP2K4 HRAS FGFR1 CDKN2A RB1 HUWE1 KRAS HRAS CLSPN RET BRAF NOTCH KRAS CTNND1 TP53 NOTCH4 NF1 HDAC4 UTX PIK3CA NF2 MAPK13 PTEN PDGFRA MDCI RB1 PDGFRB NOTCHi TP53 PIK3CA PTEN TP53 RB1 SMAD4 SMARCAL1 STAT4 TIMELESS TP53

UTX

WO 2013/075059 PCT/US2012/065724 -89 -~~r N m ~ 0 ' f e-~ - ~- F-ClP. rnA r r'~~~F 11!Cld C ~A _ ~4b, -Ca ID, m u u ~, U, -z u -~ .~ ',- 'U, WO 2013/075059 PCT/US2012/065724 -90 C)c C) M-~ 000 t-t CIAL a4 ~ - r WO 2013/075059 PCT/US2012/065724 -91 u ui u 1- WO 2013/075059 PCT/US2012/065724 -92 The teachings of all patents, published applications and references cited herein are incorporated by reference in their entirety. 5 While this invention has been particularly shown and described with references to example embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the scope of the invention encompassed by the appended claims.

Claims

1. A method of determining a triple negative breast cancer (TNBC) subtype in an individual in need thereof comprising 5 a) determining expression of one or more genes in one or more TNBC cells of the individual; and b) comparing the expression of the one or more genes in the TNBC cells with the expression of the one or more genes in a control, wherein increased expression of one or more cell cycle genes, cell 10 division genes, proliferation genes, DNA damage response genes or a combination thereof in the TNBC cells compared to the control indicates that the TNBC is a TNBC BL1 subtype; increased expression of one or more growth factor signaling genes, glycolysis genes, gluconeogenesis genes or a combination thereof in the 15 TNBC cells compared to the control indicates that the TNBC is a TNBC BL2 subtype; increased expression of one or more immune cell signaling genes, cytokine signaling genes, antigen processing and presentation genes, core immune signal transduction genes or a combination thereof in the TNBC 20 cells compared to the control indicates that the TNBC is a TNBC TM subtype; increased expression of one or more cell motility genes, extracellular matrix (ECM) receptor interaction genes, cell differentiation genes or a combination thereof in the TNBC cells compared to the control indicates that 25 the TNBC is a TNBC M subtype; increased expression of one or more cell motility genes, cellular differentiation genes, growth pathway genes, growth factor signaling genes, angiogenesis genes, immune signaling genes, stem cell genes, HOX genes, mesenchymal stem cell-specific marker genes or a combination thereof in the WO 2013/075059 PCT/US2012/065724 -94 TNBC cells compared to the control indicates that the TNBC is a TNBC MSL subtype; and increased expression of one or more hormone regulated genes in the TNBC cells compared to the control indicates that the TNBC is a TNBC 5 LAR subtype.

2. The method of Claim I wherein detection of the TNBC BLI subtype comprises selectively detecting increased expression of a gene combination comprising AURKA, AURKB, CENPA, CENPF, BUB1, BUBIB, TTK, CCNA2, 10 PLKJ, PRCI, MYC, NRAS, PLKJ, BIRC5, CHEK1, FANCA, FANCG, FOXM, HNGA1, RAD54BP, R AD51, NEKS, NBN, EXO1, MSH2, MCM10, RAD21,SIX3, Z1 C1, SOX4, SOXJ and MDC1 wherein detection of increased expression of the gene combination indicates that the TNBC is TNBC BL1 subtype, 15

3. The method of Claim 1 or 2 wherein determination that the TNBC is a TNBC BLI subtype further comprises detecting increased expression of Ki-67 mRNA in the TNBC cells compared to the control.

4. The method of Claims 1, 2 ,or 3 wherein determination that the TNBC is a 20 TNBC BL1 subtype further comprises detecting one or more mutated genes.

5. The method of Claim 4 wherein the one or more mutated genes comprise mutated BRCA1, STAT4, UTX, BRCA2, TP53, CTNND1, TOP2B, CAMK]G, MAPK13, MDC1, PTEN, RB1, SMAD4, CDKNT2A, ATM, ATR, CLSPN, HDAC4, 25 NOTCH, SMARCAL], and TIMELESS.

6. The method of Claim 1 wherein detection of the TNBC BL2 subtype comprises selectively detecting increased expression of a gene combination comprising EGFR, MET, ELF4, MAF, NUAK1, JAG], FOSL2, 1D], ZIC1, SOX11, 30 1D3, FHL2, EPHA2, TP63 and MME wherein detection of increased expression of the gene combination indicates that the TNBC is TNBC BL2 subtype, WO 2013/075059 PCT/US2012/065724 -95

7. The method of Claim 1 or 6 wherein determination that the TNBC is a TNBC BL2 subtype further comprises detecting one or more mutated genes, 5

8. The method of Claim 7 wherein the one or more mutated genes comprise mutated BRCA1, RBI, TP53, PTEN, CDKN2A, UTX, BRAC2, PTCH1, PTCH2, and RET

9. The method of Claim 1 wherein detection of the TNBC M subtype comprises 10 selectively detecting increased expression of a gene combination comprising TGFBILI, BGN, SMAD6, SMAD7, NOTCH], TGFB1, TGFB2, TGFB3, TGFBR1, TGFBR2, TGFBR3, MMP2, ACTA2, SNAI2, SPARC, SMAD7, PDGFRA, TAGLN, TCF4, TWISTS, ZEB], COL3A1, JAG], EN], MYLK, STK38L, CDH11, ETV5, IGFlR, FGFR1, FGFR2, FGFR3, TBX3, COL5A2, GNG11, ZEB2, CTNNB1, 15 DKK2, DKK3, SFPR4, TCF4, TCF7L2, FZD4, CA VI, CAV2, CCND1 and CCND2 wherein detection of increased expression of the gene combination indicates that the TNBC is TNBC M subtype,

10. The method of Claim 1 or 9 wherein determination that the TNBC is a 20 TNBC M subtype further comprises detecting decreased E-cadherin (CDH1) expression compared to the control.

11. The method of Claims 1, 9 or 10 wherein determination that the TNBC is a TNBC M subtype further comprises detecting one or more mutated genes. 25

12. The method of Claim 11 wherein the one or more mutated genes comprise mutated PTEN, RB1, TP53, PIK3CA, APC, BRAF, CTNNB1, FGFR1, GLI], HRAS, KRAS, NOTCH], and NOTCH4. WO 2013/075059 PCT/US2012/065724 -96

13. The method of Claim 1, 9, 10, 11 or 12 wherein determination that the TNBC is a TNBC M subtype further comprises detecting chromosomal amplifications in KRAS, IGFIR or MET. 5

14. The method of Claim 1 wherein detection of the TNBC MSL subtype comprises selectively detecting increased expression of a gene combination comprising VEGFR2 (KDR), TEK, TIE], EPAS1, ABCA8, PROCR, ENG, ALDHAJ, PER], ABCB1, TERF2IP, BCL2, BMP2, EPAS1, STAT4, PPARG, JAKi, ID], SMAD3, TWIST], THY1, HOXA5, HOXA10, GL13, HHEX, ZFHX4, HMBOX1, 10 FOS, PIK3RI, MF, MAFB, RPS6KA2, TCF4, TGFBILI, MEISI, MEIS2, MEOX, MEOX2, MSXI, ITGA V, KDR, NGFR, NT5E, PDGFRA, PDGFRB, POU2FJ, and VCAMI wherein detection of increased expression of the gene combination indicates that the TNBC is TNBC MSL subtype.

15 15. The method of Claim 1 or 14 wherein determination that the TNBC is a TNBC MSL subtype further comprises detecting low expression of claudins 3, 4, 7 or a combination thereof compared to the control.

16. The method of Claims 1, 14 or 15 wherein determination that the TNBC is a 20 MSL subtype TNBC further comprises detecting one or more mutated genes.

17. The method of Claim 16 wherein the one or more mutated genes comprise mutated CDKN2A, HRAS, TP53, NF1, PIK3CA, BRCA1, BRAF, KRAS, NF2, PDGFRA, APC, CTNNB1, FGFR1, PDGFRB. 25

18, The method of Claim I wherein detection of the TNBC IM subtype comprises selectively detecting increased expression of a gene combination comprising CCL19, CCL3, CCL4, CCL5, CCL8, CCR1, CCR2, CCR5, CCR7, CD2, CD3 7, CD38, CD3D, CD48, CD52, CD69, CD74, and CD8A wherein detection of 30 increased expression of the gene combination indicates that the TNBC is TNBC IM subtype. WO 2013/075059 PCT/US2012/065724 -97

19. The method of Claim 1 or 18 wherein determination that the TNBC is a TNBC IM subtype further comprises detecting one or more mutated genes. 5

20. The method of Claim 19 wherein the one or more mutated genes comprise TP53, CTNNA, DDX18, HUWE], NFKBIA, APC, BRAF, MAP2K4, RB], STAT4, STAT], and RET.

21. The method of Claim 1 wherein detection of the TNBC LAR subtype 10 comprises selectively detecting increased expression of a gene combination comprising AR, DHCR24, ALCAM GATA2, GATA3, IDIH, IDIH2, CDHI1, ERBB3, CUX2, FGFR4, HOPX, FASN, FKBP5, APOD, PIP, SPDEF, CLDN8, FOXA1, KRT1 8, and XBP1 wherein detection of increased expression of the gene combination indicates that the TNBC is TNBC LAR subtype, 15

22. The method of Claim 1 or 21 wherein determination that the TNBC is a LAR subtype TNBC further comprises detecting one or more mutated genes.

23. The method of Claim 22 wherein the one or more mutated genes comprise 20 PIK3CA, CDHI, PTEN, RB], TPS3, and MAP3KI.

24. The method of any one of Claims 2-23 wherein selective expression of the gene combination is detected using PCR, microarray analysis, next generation RNA sequencing, immunofluorescence, Western blot analysis or 25 a combination thereof.

25. The method of any one of Claims 4, 5, 7, 8, 11, 12, 16, 17, 19 or 20 wherein the one or more mutated genes is detected using a SNaPshot assay, exome sequencing, sanger gene sequencing, resequencing array analysis, mRNA 30 analysis/cDNA sequencing polymerase chain reaction (PCR), single-strand conformation polymorphism (sscp), heteroduplex analysis (het), allele- WO 2013/075059 PCT/US2012/065724 -98 specific oligonucleotide (aso), restriction fragment analysis, allele-specific amplification (asa), single nucleotide primer extension, oligonucleotide ligation assay (ola), denaturing gradient gel electrophoresis (DGGE), temperature gradient gel electrophoresis (TGGE) and single strand 5 conformation polymorphism (SSCP) or a combination thereof.

26. The method of any one of the preceding claims wherein the individual has been diagnosed with breast cancer, the individual has not been diagnosed with breast cancer, or the individual is at risk of developing breast cancer, 10

27. The method of any one of the preceding claims wherein the individual is a human.

28. The method of any preceding claim wherein the TNBC cells are epithelial 15 cells, mesenchymal cells, immune cells or a combination thereof.

29, The method of any preceding claim wherein the control is a non-cancer cell.

30. The method of Claim 29 wherein the non-cancer cell is obtained from an 20 individual that does not have TNBC, from non-cancerous tissue from an individual with TNBC or a combination thereof.

31. The method of any one of the preceding claims further comprising determining a treatment protocol for the triple negative breast cancer 25 (TNBC) patient based on the TNBC subtype.

32. The method of Claim 31 wherein the TNBC BLI or BL2 subtype is detected in the individual and the individual is treated with one or more drugs that damages DNA. 30 WO 2013/075059 PCT/US2012/065724 -99-

33. The method of Claim 32 wherein the one or more drugs is an alkylating-like agent, a PARP inhibitor or a combination thereof.

34. The method of Claim 33 wherein the alkylating-like drug is cisplatin and the 5 PARP inhibitor is veliparib, olaparib or a combination thereof.

35. The method of Claim 31 wherein the TNBC ML subtype is detected in the individual and the individual is treated with one or more drugs that inhibits Src, IGF IR, MET, PDGFR or a combination thereof. 10

36. The method of Claim 35 wherein the one or more drugs is disatinib, OSI 906, BMS-754807, PF2341066, sorafinib or a combination thereof.

37. The method of Claim 35 or 36 further comprising treating the individual with 15 a P13K inhibitor, a Pl3K/mTOR inhibitor or a combination thereof.

38. The method of 37 wherein the P13K inhibitor is BKM-120, GDC0941 or a combination thereof, and the P13K/mTOR inhibitor is NVP-BEZ23 5, GDC0980 or a combination thereof. 20

39. The method of Claim 31 wherein the TNBC LAR subtype is detected in the individual and the individual is treated with one or more drugs that inhibit androgen receptor (AR). 25

40. The method of Claim 39 wherein the one or more drugs that inhibit AR is bicalutamide, MVD3 100, abiraterone or a combination thereof.

41. The method of Claim 39 or 40 further comprising treating the individual with a P13K inhibitor a P13K/mTOR inhibitor or a combination thereof. 30 WO 2013/075059 PCT/US2012/065724 -100

42. The method of 41 wherein the P13K inhibitor is BKM-120, GDC0941 or a combination thereof, and the P13K/mTOR inhibitor is NVP-BEZ235, GDC0980 or a combination thereof. 5

43, The method of Claim 39, 40, 41 or 42 further comprising treating the individual with an inhibitor of HSP90,

44. The method of Claim 43 wherein the inhibitor of HSP90 is DMAG, 10

45. A method of determining whether an agent can be used to treat a triple negative breast cancer (TNBC) subtype comprising a) contacting a cell line that is a model for the TNBC subtype of interest with an agent to be assessed; and b) determining viability of the cell line in the presence of the agent, 15 wherein if the viability of the cell line decreases in the presence of the agent, then the agent can be used to treat the TNBC subtype of interest.

46. The method of Claim 45 wherein the TNBC subtype of interest is a TNBC BL-1 subtype, a TNBC BL-2 subtype, a TNBC IM subtype, a TNBC M subtype, a 20 TNBC MSL subtype or a TNBC LAR subtype.

47. A method of determining whether an agent can be used to treat a triple negative breast cancer (TNBC) BL-1 subtype comprising a) contacting one or more TNBC BL-1 subtype cell lines with an agent 25 to be assessed; and b) determining viability of the one or more cell lines in the presence of the agent, wherein if the viability of the one or more cell lines decreases in the presence of the agent, then the agent can be used to treat the TNBC BL-1 subtype. 30 WO 2013/075059 PCT/US2012/065724 -101

48. The method of Claim 47 wherein the one or more TNBC BL-1 subtype cell lines comprises HCC2157, HCC1599, HCC1937, HCC1 143, HCC3153, MDA-MB 468, and HCC38. 5

49. A method of determining whether an agent can be used to treat a triple negative breast cancer (TNBC) BL-2 subtype comprising a) contacting one or more TNBC BL-2 subtype cell lines with an agent to be assessed; and b) determining viability of the one or more cell lines in the presence of 10 the agent, wherein if the viability of the one or more cell lines decreases in the presence of the agent, then the agent can be used to treat the TNBC BL-2 subtype.

50. The method of Claim 49 wherein the one or more TNBC BL-2 subtype cell 15 lines comprises SUM149PT, CAL-851, HCC70, HCC1806 and HDQ-Pl.

51. A method of determining whether an agent can be used to treat a triple negative breast cancer (TNBC) IM subtype comprising a) contacting one or more TNBC IM subtype cell lines with an agent to 20 be assessed; and b) determining viability of the one or more cell lines in the presence of the agent, wherein if the viability of the one or more cell lines decreases in the presence of the agent, then the agent can be used to treat the TNBC IM subtype. 25

52. The method of Claim 51 wherein the one or more TNBC IM subtype cell lines comprises HCCI187 and DU4475,

53. . A method of determining whether an agent can be used to treat a triple 30 negative breast cancer (TNBC) M subtype comprising WO 2013/075059 PCT/US2012/065724 -102 a) contacting one or more TNBC M subtype cell lines with an agent to be assessed; and b) determining viability of the one or more cell lines in the presence of the agent, 5 wherein if the viability of the one or more cell lines decreases in the presence of the agent, then the agent can be used to treat the TNBC M subtype,

54. The method of Claim 53 wherein the one or more TNBC M subtype cell lines comprises BT-549, CAL-51 and CAL-120. 10

55. A method of determining whether an agent can be used to treat a triple negative breast cancer (TNBC) MSL subtype comprising a) contacting one or more TNBC MSL subtype cell lines with an agent to be assessed; and 15 b) determining viability of the one or more cell lines in the presence of the agent, wherein if the viability of the one or more cell lines decreases in the presence of the agent, then the agent can be used to treat the TNBC MSL subtype. 20

56. The method of Claim 55 wherein the one or more TNBC MSL subtype cell lines comprises HS578T, MDA-MB-157, SUM159PT, MDA-MB-436, and MDA MB-231.

57. A method of determining whether an agent can be used to treat a triple 25 negative breast cancer (TNBC) LAR subtype comprising a) contacting one or more TNBC LAR subtype cell lines with an agent to be assessed; and b) determining viability of the one or more cell lines in the presence of the agent, 30 wherein if the viability of the one or more cell lines decreases in the presence of the agent, then the agent can be used to treat the TNBC LAR subtype. WO 2013/075059 PCT/US2012/065724 -103

58. The method of Claim 57 wherein the one or more TNBC LAR subtype cell lines comprises MDA-MB, SUM185PE, HCC2185, CAL-148, and MFM-223. 5

59. The method of any one of Claims 45-58 wherein the viability is determined using a membrane leakage assay, a mitochondrial activity assay, a functional assay, a proteomic assay, a genomic assay or a combination thereof.

60. The method of any one of Claims 45-59 further comprising comparing the 10 viability of the one or more cell lines in the presence of the agent to the viability of a control.

61. The method of Claim 60 wherein the control is a non-cancerous cell line. 15

62. The method of Claim 61 wherein the control is a culture of normal human and immortalized MCF21 OA.